{ "schema_version": 1, "verified_on": "2026-10-13", "records": [ { "id": "still-measured-crop-off-3", "series": "still", "run": "measured-crop-off-3", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 117.188, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "92442a8c1dd7371cb6e1b98ebb3c6b1a543f74b1c585be65ecc0ad145101a614", "video": "/benchmarks/still-measured-crop-off-3.mp4", "poster": "/benchmarks/still-measured-crop-off-3.jpg", "api": "/benchmarks/still-measured-crop-off-3-api.json", "history": "/benchmarks/still-measured-crop-off-3-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-off-3-preview.mp4", "preview_sha256": "ef978001a58f15515e3be197a02b3cdb7848de357684c89ecae149f2d1a657af", "preview_bytes": 332149, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2652182 }, { "id": "still-measured-crop-on-4", "series": "still", "run": "measured-crop-on-4", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 118.32, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "83fab07207f85ec751858b7d3fabd90ef102792d7858c23791a98a8d7773e9e7", "video": "/benchmarks/still-measured-crop-on-4.mp4", "poster": "/benchmarks/still-measured-crop-on-4.jpg", "api": "/benchmarks/still-measured-crop-on-4-api.json", "history": "/benchmarks/still-measured-crop-on-4-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-on-4-preview.mp4", "preview_sha256": "fb4b543676ca8b384e2f00aa383d1b4c1e3589f93aa8aff263fe265f599a7929", "preview_bytes": 290530, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2321987 }, { "id": "still-measured-crop-off-5", "series": "still", "run": "measured-crop-off-5", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 118.661, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "b80897fa63a3f7fd156f597a9801731ae05897f5fe76984125fe0a72d7efc14f", "video": "/benchmarks/still-measured-crop-off-5.mp4", "poster": "/benchmarks/still-measured-crop-off-5.jpg", "api": "/benchmarks/still-measured-crop-off-5-api.json", "history": "/benchmarks/still-measured-crop-off-5-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-off-5-preview.mp4", "preview_sha256": "617f3436f0665852f68a6f54db8ffda45018c808b36922086285c79351a2a0bb", "preview_bytes": 315138, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2526441 }, { "id": "still-measured-crop-on-6", "series": "still", "run": "measured-crop-on-6", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 117.633, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "76f0a17431f43592a7c06eded9437decbf1861e551c95919d061d48c347ac5d3", "video": "/benchmarks/still-measured-crop-on-6.mp4", "poster": "/benchmarks/still-measured-crop-on-6.jpg", "api": "/benchmarks/still-measured-crop-on-6-api.json", "history": "/benchmarks/still-measured-crop-on-6-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-on-6-preview.mp4", "preview_sha256": "eeb2b09ca16a8dbff1a0ceab62f48e05e8a4f0a2a405e490259e4a0b92363b17", "preview_bytes": 290292, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2354661 }, { "id": "still-measured-crop-off-7", "series": "still", "run": "measured-crop-off-7", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 117.397, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "a3517bc75be96bdd8bba3ef2b004d31dd027ef12d45e1309dd478e68411134d1", "video": "/benchmarks/still-measured-crop-off-7.mp4", "poster": "/benchmarks/still-measured-crop-off-7.jpg", "api": "/benchmarks/still-measured-crop-off-7-api.json", "history": "/benchmarks/still-measured-crop-off-7-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-off-7-preview.mp4", "preview_sha256": "546fc909ea2f2aa7bc7cf51e7a0556f5cdf6ac38e2b002cef6cc5c42b0450955", "preview_bytes": 309038, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2453613 }, { "id": "still-measured-crop-on-8", "series": "still", "run": "measured-crop-on-8", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 117.677, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "7fec8a247bc26650671e36373de3029af37cdc4732613009b186c257187189e0", "video": "/benchmarks/still-measured-crop-on-8.mp4", "poster": "/benchmarks/still-measured-crop-on-8.jpg", "api": "/benchmarks/still-measured-crop-on-8-api.json", "history": "/benchmarks/still-measured-crop-on-8-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "In this clip, OFF adds small head turns, nods and shifts. ON keeps the head near the source pose while speech animation remains visible. These are visual observations, not a scored quality or lip-sync test.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/still-measured-crop-on-8-preview.mp4", "preview_sha256": "9485c5d31009b38ed3e6769d3025d19382da5cd9aa4e5ff14376968f27104d17", "preview_bytes": 286321, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2313761 }, { "id": "still-supplemental-full-off-9", "series": "still", "run": "supplemental-full-off-9", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "full", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 217.614, "output_seconds": 11.24, "dimensions": [ 2240, 2080 ], "sha256": "d53343a42861a9e566a16815e3f5d7d62b9abee329da56fac62d275def43e190", "video": "/benchmarks/still-supplemental-full-off-9.mp4", "poster": "/benchmarks/still-supplemental-full-off-9.jpg", "api": "/benchmarks/still-supplemental-full-off-9-api.json", "history": "/benchmarks/still-supplemental-full-off-9-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "Both outputs retain the jacket and surrounding frame. OFF adds head shifts within that framing; ON keeps the head aligned more steadily while the mouth continues animating. Review the jaw and neck boundary before using a full-image render.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Supplemental", "preview_video": "/benchmarks/still-supplemental-full-off-9-preview.mp4", "preview_sha256": "db6fd06a7dd325b2b4db3460140494e1ac09aa3aae0417468bf75fb0dbe3ed6e", "preview_bytes": 224621, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 6428114 }, { "id": "still-supplemental-full-on-10", "series": "still", "run": "supplemental-full-on-10", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "full", "size": 512, "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 218.61, "output_seconds": 11.24, "dimensions": [ 2240, 2080 ], "sha256": "441b7e0bc2fe3ddb3f84df5151c8a2f72681017663217e0b3daf7135ea0fbeff", "video": "/benchmarks/still-supplemental-full-on-10.mp4", "poster": "/benchmarks/still-supplemental-full-on-10.jpg", "api": "/benchmarks/still-supplemental-full-on-10-api.json", "history": "/benchmarks/still-supplemental-full-on-10-history.json", "source": "/sadtalker-still-mode/benchmark.json", "notes": "Both outputs retain the jacket and surrounding frame. OFF adds head shifts within that framing; ON keeps the head aligned more steadily while the mouth continues animating. Review the jaw and neck boundary before using a full-image render.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Supplemental", "preview_video": "/benchmarks/still-supplemental-full-on-10-preview.mp4", "preview_sha256": "520aead4c2b869bbb3ed862d68dfe66e2ea479295553d561f2cab6061bb10670", "preview_bytes": 188034, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 5673593 }, { "id": "preprocess-measured-full-off-4", "series": "preprocess", "run": "measured-full-off-4", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "full", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 218.925, "output_seconds": 11.24, "dimensions": [ 2240, 2080 ], "sha256": "7f3482bb34a5989f372911b4314e317de9ffc48b0ef4f3d513a03b02f78ed286", "video": "/benchmarks/preprocess-measured-full-off-4.mp4", "poster": "/benchmarks/preprocess-measured-full-off-4.jpg", "api": "/benchmarks/preprocess-measured-full-off-4-api.json", "history": "/benchmarks/preprocess-measured-full-off-4-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The jacket and original surrounding frame remain visible. The animated face moves within that scene; inspect the jaw and neck transition as the head turns. The larger final image does not mean the face model rendered more detail than the other modes.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-full-off-4-preview.mp4", "preview_sha256": "6693b2c2a5bd1cfb883e07740eab3fb66bf9244f31590c3d22891b9b6a13c529", "preview_bytes": 226467, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 6479929 }, { "id": "preprocess-measured-crop-off-5", "series": "preprocess", "run": "measured-crop-off-5", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 117.341, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "1a3299759ad92f81c82e59a09f55a82904966ad039ab44de2a0c960373590d4f", "video": "/benchmarks/preprocess-measured-crop-off-5.mp4", "poster": "/benchmarks/preprocess-measured-crop-off-5.jpg", "api": "/benchmarks/preprocess-measured-crop-off-5-api.json", "history": "/benchmarks/preprocess-measured-crop-off-5-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The automatic crop produces a square close-up of the face, removing most shoulders and the wider source composition. Mouth, eyes and head movement remain visible. At the same display width the face appears larger, which also makes small mouth and edge artifacts easier to inspect.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-crop-off-5-preview.mp4", "preview_sha256": "ea37287b428fb8609ee0e10f7997f7c1948ae941a317885e27107d0db1378b63", "preview_bytes": 314921, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2505313 }, { "id": "preprocess-measured-resize-off-6", "series": "preprocess", "run": "measured-resize-off-6", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "resize", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 116.527, "output_seconds": 11.24, "dimensions": [ 1024, 960 ], "sha256": "b6aec76dc7a9f829571468de6b5ee49146ba6f26ae43eb164f84af612e95571b", "video": "/benchmarks/preprocess-measured-resize-off-6.mp4", "poster": "/benchmarks/preprocess-measured-resize-off-6.jpg", "api": "/benchmarks/preprocess-measured-resize-off-6-api.json", "history": "/benchmarks/preprocess-measured-resize-off-6-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The output retains the wider portrait without pasting the face back into a static source frame. The face and nearby collar/shoulder area are rendered as part of that image. It completed on this portrait; that does not establish that resize will work well when a face occupies a much smaller part of the source.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-resize-off-6-preview.mp4", "preview_sha256": "3799050bfa88e579ab1a603e40732fb41ad52a43ee42e9acef83ef005fa5a021", "preview_bytes": 286245, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 1876574 }, { "id": "preprocess-measured-crop-off-7", "series": "preprocess", "run": "measured-crop-off-7", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 120.079, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "099c7df34c47245c574173020e71ccfdafa140bd3cad60845872a0db39519f43", "video": "/benchmarks/preprocess-measured-crop-off-7.mp4", "poster": "/benchmarks/preprocess-measured-crop-off-7.jpg", "api": "/benchmarks/preprocess-measured-crop-off-7-api.json", "history": "/benchmarks/preprocess-measured-crop-off-7-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The automatic crop produces a square close-up of the face, removing most shoulders and the wider source composition. Mouth, eyes and head movement remain visible. At the same display width the face appears larger, which also makes small mouth and edge artifacts easier to inspect.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-crop-off-7-preview.mp4", "preview_sha256": "3df09034bb6dfeaff370aff368b16b05bc5a67453efddd1154655c611fa96f00", "preview_bytes": 337019, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2636809 }, { "id": "preprocess-measured-resize-off-8", "series": "preprocess", "run": "measured-resize-off-8", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "resize", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 116.053, "output_seconds": 11.24, "dimensions": [ 1024, 960 ], "sha256": "7d1bc3b2f285245a8323a84db2a2b5507571daa8af0896d26b3c069cac65842e", "video": "/benchmarks/preprocess-measured-resize-off-8.mp4", "poster": "/benchmarks/preprocess-measured-resize-off-8.jpg", "api": "/benchmarks/preprocess-measured-resize-off-8-api.json", "history": "/benchmarks/preprocess-measured-resize-off-8-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The output retains the wider portrait without pasting the face back into a static source frame. The face and nearby collar/shoulder area are rendered as part of that image. It completed on this portrait; that does not establish that resize will work well when a face occupies a much smaller part of the source.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-resize-off-8-preview.mp4", "preview_sha256": "de9172787e252230fecff998efb9d1a2f0e4beee778eda726136f60ffcce3c4b", "preview_bytes": 283947, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 1881098 }, { "id": "preprocess-measured-full-off-9", "series": "preprocess", "run": "measured-full-off-9", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "full", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 217.212, "output_seconds": 11.24, "dimensions": [ 2240, 2080 ], "sha256": "d965de7fdefcac974fc40e267e27cb9f051deb1857924785dcb9d1029d735658", "video": "/benchmarks/preprocess-measured-full-off-9.mp4", "poster": "/benchmarks/preprocess-measured-full-off-9.jpg", "api": "/benchmarks/preprocess-measured-full-off-9-api.json", "history": "/benchmarks/preprocess-measured-full-off-9-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The jacket and original surrounding frame remain visible. The animated face moves within that scene; inspect the jaw and neck transition as the head turns. The larger final image does not mean the face model rendered more detail than the other modes.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-full-off-9-preview.mp4", "preview_sha256": "4032af23eb02179646d965c5d1656ca2b149914f60b493d8e2e8599fe0100b05", "preview_bytes": 211291, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 6260139 }, { "id": "preprocess-measured-resize-off-10", "series": "preprocess", "run": "measured-resize-off-10", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "resize", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 116.08, "output_seconds": 11.24, "dimensions": [ 1024, 960 ], "sha256": "27403086e2d3feaeb4a8c0f987ce8c3fd5c92bab7aa6efbbb0e33b26a9e9891a", "video": "/benchmarks/preprocess-measured-resize-off-10.mp4", "poster": "/benchmarks/preprocess-measured-resize-off-10.jpg", "api": "/benchmarks/preprocess-measured-resize-off-10-api.json", "history": "/benchmarks/preprocess-measured-resize-off-10-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The output retains the wider portrait without pasting the face back into a static source frame. The face and nearby collar/shoulder area are rendered as part of that image. It completed on this portrait; that does not establish that resize will work well when a face occupies a much smaller part of the source.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-resize-off-10-preview.mp4", "preview_sha256": "751f0995a1a2d93f5d942103496e3a1def8938c94236b3d26e16595c8c717ff1", "preview_bytes": 274426, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 1770014 }, { "id": "preprocess-measured-full-off-11", "series": "preprocess", "run": "measured-full-off-11", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "full", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 218.552, "output_seconds": 11.24, "dimensions": [ 2240, 2080 ], "sha256": "0d1411c5580045b2e7697f8ff3f24bfc2c270093fc5f1f32b39bd3cd544daaf9", "video": "/benchmarks/preprocess-measured-full-off-11.mp4", "poster": "/benchmarks/preprocess-measured-full-off-11.jpg", "api": "/benchmarks/preprocess-measured-full-off-11-api.json", "history": "/benchmarks/preprocess-measured-full-off-11-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The jacket and original surrounding frame remain visible. The animated face moves within that scene; inspect the jaw and neck transition as the head turns. The larger final image does not mean the face model rendered more detail than the other modes.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-full-off-11-preview.mp4", "preview_sha256": "cf877678e70bc587387eb7b6e6e05665ead70c0caca4f8f60217ce402aab01c0", "preview_bytes": 201903, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 5996239 }, { "id": "preprocess-measured-crop-off-12", "series": "preprocess", "run": "measured-crop-off-12", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 118.131, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "92fddebf5f5fad5ff2e5fbca8b183a8bf4ba08198168917e59ad438397d98ce1", "video": "/benchmarks/preprocess-measured-crop-off-12.mp4", "poster": "/benchmarks/preprocess-measured-crop-off-12.jpg", "api": "/benchmarks/preprocess-measured-crop-off-12-api.json", "history": "/benchmarks/preprocess-measured-crop-off-12-history.json", "source": "/sadtalker-preprocess-comparison/benchmark.json", "notes": "The automatic crop produces a square close-up of the face, removing most shoulders and the wider source composition. Mouth, eyes and head movement remain visible. At the same display width the face appears larger, which also makes small mouth and edge artifacts easier to inspect.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Repeat trial", "preview_video": "/benchmarks/preprocess-measured-crop-off-12-preview.mp4", "preview_sha256": "0e55a61394d46dcd60f910cdcfa051ffb070e78f145cbb4024abb1df6b1ca92d", "preview_bytes": 318484, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2507410 }, { "id": "quality-256-off", "series": "quality", "run": "256-off", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 256, "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 27.883, "output_seconds": 11.24, "dimensions": [ 256, 256 ], "sha256": "6a4e5f25a88dc9289a288a58567fb96dc95ebcde0d7152d7169a658a59fe469d", "video": "/benchmarks/quality-256-off.mp4", "poster": "/benchmarks/quality-256-off.jpg", "api": "/benchmarks/quality-256-off-api.json", "history": "/benchmarks/quality-256-off-history.json", "source": "/sadtalker-quality-tests/benchmark.json", "notes": "In the sampled close-ups, the 256 render looks softer around the eyes, beard and skin texture. The 512 render shows finer-looking facial texture at this enlarged display size. Mouth shape, blinking and head position also differ between the independent runs, so those differences cannot be attributed to resolution alone. Neither sample establishes objectively better lip sync or identity accuracy.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Single configuration run", "preview_video": "/benchmarks/quality-256-off-preview.mp4", "preview_sha256": "3e6f108289bfee586bbc9b0b3701878398e9e766704fe3ba9379a5b204aa3a31", "preview_bytes": 178630, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 379336 }, { "id": "quality-512-off", "series": "quality", "run": "512-off", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 55.405, "output_seconds": 11.24, "dimensions": [ 512, 512 ], "sha256": "24a267efa757ede455c55d5f07ba98ab74ba4c67db080b9e6460b187324bdfc4", "video": "/benchmarks/quality-512-off.mp4", "poster": "/benchmarks/quality-512-off.jpg", "api": "/benchmarks/quality-512-off-api.json", "history": "/benchmarks/quality-512-off-history.json", "source": "/sadtalker-quality-tests/benchmark.json", "notes": "In the sampled close-ups, the 256 render looks softer around the eyes, beard and skin texture. The 512 render shows finer-looking facial texture at this enlarged display size. Mouth shape, blinking and head position also differ between the independent runs, so those differences cannot be attributed to resolution alone. Neither sample establishes objectively better lip sync or identity accuracy.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Single configuration run", "preview_video": "/benchmarks/quality-512-off-preview.mp4", "preview_sha256": "05d9c426761b77ae59967a15ffa0c56a1c73d2ec1cc36f59481a5d19fb3a5eaa", "preview_bytes": 401955, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 1062498 }, { "id": "quality-256-on", "series": "quality", "run": "256-on", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 256, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 77.216, "output_seconds": 11.24, "dimensions": [ 512, 512 ], "sha256": "fa4835cf2668723852be2f0a5513b5f239cbd4e150eda23c64224024fbb02e40", "video": "/benchmarks/quality-256-on.mp4", "poster": "/benchmarks/quality-256-on.jpg", "api": "/benchmarks/quality-256-on-api.json", "history": "/benchmarks/quality-256-on-history.json", "source": "/sadtalker-quality-tests/benchmark.json", "notes": "In the sampled close-ups, GFPGAN ON gives eyes and facial hair more defined edges and produces smoother-looking skin than OFF. The mouth and teeth also look different, but motion and expression differ between these independent renders. The sharper presentation is visible here; exact identity recovery and improved temporal stability were not measured.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Single configuration run", "preview_video": "/benchmarks/quality-256-on-preview.mp4", "preview_sha256": "39385fb109d09404e9b31785b1b90bfa7443a46a13f459738bacbca38eaad33d", "preview_bytes": 394748, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 832997 }, { "id": "quality-512-on", "series": "quality", "run": "512-on", "date": "2026-10-04", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/demo-media/sadtalker-corporate-presenter-source.png", "audio": "/demo-media/sadtalker-male-narration-audio.wav", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "f133f9117a71f9783c9588635a632d6a56f85d258ba6dceb1371d8f51f021e48" }, "settings": { "preprocess": "crop", "size": 512, "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 119.392, "output_seconds": 11.24, "dimensions": [ 1024, 1024 ], "sha256": "9c98bf5f114c925524aeb76f7a7ba39c48d9ef49ce9382f160b32bc141b1b937", "video": "/benchmarks/quality-512-on.mp4", "poster": "/benchmarks/quality-512-on.jpg", "api": "/benchmarks/quality-512-on-api.json", "history": "/benchmarks/quality-512-on-history.json", "source": "/sadtalker-quality-tests/benchmark.json", "notes": "In the sampled close-ups, GFPGAN ON gives eyes and facial hair more defined edges and produces smoother-looking skin than OFF. The mouth and teeth also look different, but motion and expression differ between these independent renders. The sharper presentation is visible here; exact identity recovery and improved temporal stability were not measured.", "notes_scope": "Series observations from the source comparison; not an individual quality score.", "trial_kind": "Single configuration run", "preview_video": "/benchmarks/quality-512-on-preview.mp4", "preview_sha256": "56f7d3eb65f2a06ebc426b53d5924ffe27d336c657decb5aa1742ff5b08c1df4", "preview_bytes": 348380, "audio_style": "Male narration", "input_title": "Studio presenter", "clip_description": "Original narration from the controlled settings comparisons.", "video_bytes": 2660800 }, { "id": "new-01-lincoln", "series": "variants", "run": "new-01-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.742, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "5cc9930560b424fe91c4c652db012897e7d3bec9e1a0d68d798d29337d1a9e1f", "video": "/benchmarks/new-01-lincoln.mp4", "poster": "/benchmarks/new-01-lincoln.jpg", "api": "/benchmarks/new-01-lincoln-api.json", "history": "/benchmarks/new-01-lincoln-history.json", "source": "/benchmarks/new-01-lincoln-source.json", "source_metadata": "/benchmarks/new-01-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.65, "mean_power_w": 81.1055 }, "poster_seconds": 2, "notes": "The sampled face crop retains the scan\u2019s sepia tone and soft texture. Teeth and the open-mouth edge appear less uniform than the closed mouth in the original scan.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 71406, "preview_video": "/benchmarks/new-01-lincoln-preview.mp4", "preview_sha256": "f1b8f1c0edbb71358fbb0a5df618c01315fc877233daac87ed3d30815167f7d3", "preview_bytes": 46844, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-tubman", "series": "variants", "run": "new-01-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.619, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "82a69c099c303e8724f668b2f33628c837027e3bbb07aac44faca8ab070b24a6", "video": "/benchmarks/new-01-tubman.mp4", "poster": "/benchmarks/new-01-tubman.jpg", "api": "/benchmarks/new-01-tubman-api.json", "history": "/benchmarks/new-01-tubman-history.json", "source": "/benchmarks/new-01-tubman-source.json", "source_metadata": "/benchmarks/new-01-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 10.25, "mean_power_w": 88.392 }, "poster_seconds": 2, "notes": "The sampled crop enlarges the archival face and shows an open mouth. Facial edges remain soft; the original card border and seated body are outside this crop.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 78365, "preview_video": "/benchmarks/new-01-tubman-preview.mp4", "preview_sha256": "bb3d92f8074a42d34d0f30653414dcdebff2101015954d4bee145dafd4fc1bc4", "preview_bytes": 49527, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-ai-woman", "series": "variants", "run": "new-01-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.377, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "9ad21579afca2479fb587995fe359fc84b74e0ca347eba6adc73e0891bf12eb1", "video": "/benchmarks/new-01-ai-woman.mp4", "poster": "/benchmarks/new-01-ai-woman.jpg", "api": "/benchmarks/new-01-ai-woman-api.json", "history": "/benchmarks/new-01-ai-woman-history.json", "source": "/benchmarks/new-01-ai-woman-source.json", "source_metadata": "/benchmarks/new-01-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 10.047619047619047, "mean_power_w": 83.75190476190477 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 109074, "preview_video": "/benchmarks/new-01-ai-woman-preview.mp4", "preview_sha256": "d96a10f23155c2a3a47e351661c776154aa57ba2dcfcfae1c389c8b70592f8ba", "preview_bytes": 58893, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-corporate", "series": "variants", "run": "new-01-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.023, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "6dc929e35c32a5806f3f7cef31b72c7bfc7a823024b21d1d40f4fe45f9abf427", "video": "/benchmarks/new-01-corporate.mp4", "poster": "/benchmarks/new-01-corporate.jpg", "api": "/benchmarks/new-01-corporate-api.json", "history": "/benchmarks/new-01-corporate-history.json", "source": "/benchmarks/new-01-corporate-source.json", "source_metadata": "/benchmarks/new-01-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.523809523809524, "mean_power_w": 90.02142857142857 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 154955, "preview_video": "/benchmarks/new-01-corporate-preview.mp4", "preview_sha256": "fd7f2e70593f45dd922d23d6b59c3e320dba204a02bd36ccb04416efe70f8138", "preview_bytes": 71671, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-female", "series": "variants", "run": "new-01-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.013, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "23737ccfd2bad2e259ffcd7cdcfe4e8fee6a0fbfefd42107232bc567c5ea35fc", "video": "/benchmarks/new-01-female.mp4", "poster": "/benchmarks/new-01-female.jpg", "api": "/benchmarks/new-01-female-api.json", "history": "/benchmarks/new-01-female-history.json", "source": "/benchmarks/new-01-female-source.json", "source_metadata": "/benchmarks/new-01-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.476190476190476, "mean_power_w": 91.15904761904763 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth become visible during speech; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 152000, "preview_video": "/benchmarks/new-01-female-preview.mp4", "preview_sha256": "af97a885e99b3492113b53c570f2dcf21a02acd53a74581d86a904d724b3561d", "preview_bytes": 75255, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-american", "series": "variants", "run": "new-01-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.077, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "7e238d5001fb9e4425acf498596b4147331423f328a6ad7f9f915a523d48d366", "video": "/benchmarks/new-01-american.mp4", "poster": "/benchmarks/new-01-american.jpg", "api": "/benchmarks/new-01-american-api.json", "history": "/benchmarks/new-01-american-history.json", "source": "/benchmarks/new-01-american-source.json", "source_metadata": "/benchmarks/new-01-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4091.0, "mean_utilization_pct": 10.666666666666666, "mean_power_w": 89.28333333333333 }, "poster_seconds": 2, "notes": "The sampled frame retains warm background lighting. Mouth shape and visible teeth differ from the closed-mouth source; compare facial detail at the intended display size.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 130169, "preview_video": "/benchmarks/new-01-american-preview.mp4", "preview_sha256": "b7ef39dc18a600a393d4a4e4dd8f25ad064e9cf894f9cede7e3622474c328b9f", "preview_bytes": 67498, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-young", "series": "variants", "run": "new-01-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.931, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "ab47ab32f08b62f12634890bd1d5ec4f9b226922ae5fabd2412db3ba3f362cde", "video": "/benchmarks/new-01-young.mp4", "poster": "/benchmarks/new-01-young.jpg", "api": "/benchmarks/new-01-young-api.json", "history": "/benchmarks/new-01-young-history.json", "source": "/benchmarks/new-01-young-source.json", "source_metadata": "/benchmarks/new-01-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.380952380952381, "mean_power_w": 92.70238095238095 }, "poster_seconds": 2, "notes": "The sampled frame retains the source\u2019s shadowed lighting and long hair. The mouth is open while parts of the eyes and cheeks remain dark, making fine facial detail harder to inspect.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 111923, "preview_video": "/benchmarks/new-01-young-preview.mp4", "preview_sha256": "423fe04c415598d071083b9719a649ff7a4f1277f8caac2a63f5c35eda5695ad", "preview_bytes": 59362, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-01-avatar", "series": "variants", "run": "new-01-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.203, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "17cdfdbb42e4ae8742a84fb34e54d912c4518c92ccbda40bb1ddea2d88713b42", "video": "/benchmarks/new-01-avatar.mp4", "poster": "/benchmarks/new-01-avatar.jpg", "api": "/benchmarks/new-01-avatar-api.json", "history": "/benchmarks/new-01-avatar-history.json", "source": "/benchmarks/new-01-avatar-source.json", "source_metadata": "/benchmarks/new-01-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3995.0, "mean_utilization_pct": 11.714285714285714, "mean_power_w": 82.62333333333333 }, "poster_seconds": 2, "notes": "The sampled frame retains the outdoor background and textured hair. Mouth opening and eye position differ from the source; inspect the jaw and neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 97962, "preview_video": "/benchmarks/new-01-avatar-preview.mp4", "preview_sha256": "46e03f5a6837729d79789ee81bc57408bd85137f104af562f85d1ee0e286b74e", "preview_bytes": 56598, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-lincoln", "series": "variants", "run": "new-02-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.501, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "3aa74fe978d3e10a9acbda2bf5f02bc66b2ca5d13106342fb3924816728deb23", "video": "/benchmarks/new-02-lincoln.mp4", "poster": "/benchmarks/new-02-lincoln.jpg", "api": "/benchmarks/new-02-lincoln-api.json", "history": "/benchmarks/new-02-lincoln-history.json", "source": "/benchmarks/new-02-lincoln-source.json", "source_metadata": "/benchmarks/new-02-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8859.0, "mean_utilization_pct": 40.06666666666667, "mean_power_w": 170.04366666666667 }, "poster_seconds": 2, "notes": "The sampled face crop retains the scan\u2019s sepia tone and soft texture. Teeth and the open-mouth edge appear less uniform than the closed mouth in the original scan.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 117932, "preview_video": "/benchmarks/new-02-lincoln-preview.mp4", "preview_sha256": "fbc0296978e1beb75544c4c6658b08ed3f1a91672f1cdba2557c512d97b030f0", "preview_bytes": 73587, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-tubman", "series": "variants", "run": "new-02-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.425, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "5377a4473854954286559abd390c9ed4bdb56bea6ca6a2742febfb3d69a5444b", "video": "/benchmarks/new-02-tubman.mp4", "poster": "/benchmarks/new-02-tubman.jpg", "api": "/benchmarks/new-02-tubman-api.json", "history": "/benchmarks/new-02-tubman-history.json", "source": "/benchmarks/new-02-tubman-source.json", "source_metadata": "/benchmarks/new-02-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 36.8, "mean_power_w": 171.90166666666667 }, "poster_seconds": 2, "notes": "The sampled crop enlarges the archival face and shows an open mouth. Facial edges remain soft; the original card border and seated body are outside this crop.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 138854, "preview_video": "/benchmarks/new-02-tubman-preview.mp4", "preview_sha256": "6d905076ec8e118b4b95a7248ffc7b1b5628e3f96765b14c3ef84d8f9e850fd3", "preview_bytes": 85131, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-ai-woman", "series": "variants", "run": "new-02-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.371, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "43ba0129fdc3089d839e282ba60aa27f570d8bf318e746f98fef29cc5adaeebe", "video": "/benchmarks/new-02-ai-woman.mp4", "poster": "/benchmarks/new-02-ai-woman.jpg", "api": "/benchmarks/new-02-ai-woman-api.json", "history": "/benchmarks/new-02-ai-woman-history.json", "source": "/benchmarks/new-02-ai-woman-source.json", "source_metadata": "/benchmarks/new-02-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8955.0, "mean_utilization_pct": 40.903225806451616, "mean_power_w": 171.95580645161291 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 224381, "preview_video": "/benchmarks/new-02-ai-woman-preview.mp4", "preview_sha256": "a58baa5440cd6795e7f5f3db6d0e0969d3b901f8f506eb8785d8646ae7a07d11", "preview_bytes": 111667, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-corporate", "series": "variants", "run": "new-02-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.972, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "b2c168e706859b08bb7e36fe595565d8bcbae8c90f9b962ef4af23616f11f147", "video": "/benchmarks/new-02-corporate.mp4", "poster": "/benchmarks/new-02-corporate.jpg", "api": "/benchmarks/new-02-corporate-api.json", "history": "/benchmarks/new-02-corporate-history.json", "source": "/benchmarks/new-02-corporate-source.json", "source_metadata": "/benchmarks/new-02-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 42.86666666666667, "mean_power_w": 175.91666666666666 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 338410, "preview_video": "/benchmarks/new-02-corporate-preview.mp4", "preview_sha256": "fbbe2fa727ea126d7f46a86e0779b62af2760d313e08211a978e8b5c6e126084", "preview_bytes": 134381, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-female", "series": "variants", "run": "new-02-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.929, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "d5959bc14ca4cfc34bce7bd78f73ef916dc6785b0502e4a7fbd65fdcb47dc218", "video": "/benchmarks/new-02-female.mp4", "poster": "/benchmarks/new-02-female.jpg", "api": "/benchmarks/new-02-female-api.json", "history": "/benchmarks/new-02-female-history.json", "source": "/benchmarks/new-02-female-source.json", "source_metadata": "/benchmarks/new-02-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 40.13333333333333, "mean_power_w": 176.58133333333333 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth become visible during speech; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 259747, "preview_video": "/benchmarks/new-02-female-preview.mp4", "preview_sha256": "6052b4528b8f6d7de7ab3e8be936ca8ab541165dce2c03a00e09b61a846b772d", "preview_bytes": 121660, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-american", "series": "variants", "run": "new-02-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.103, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "a034d5b9ddcb9d6711173fc6056fa7691be5ae34d1aec4a042e34e45c376bf2c", "video": "/benchmarks/new-02-american.mp4", "poster": "/benchmarks/new-02-american.jpg", "api": "/benchmarks/new-02-american-api.json", "history": "/benchmarks/new-02-american-history.json", "source": "/benchmarks/new-02-american-source.json", "source_metadata": "/benchmarks/new-02-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 38.74193548387097, "mean_power_w": 174.51096774193547 }, "poster_seconds": 2, "notes": "The sampled frame retains warm background lighting. Mouth shape and visible teeth differ from the closed-mouth source; compare facial detail at the intended display size.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 299650, "preview_video": "/benchmarks/new-02-american-preview.mp4", "preview_sha256": "90351a51af8c1bfb3a1ed30e3f83a4e235baf1fc22589e3acd88e6d2cd0bbe5d", "preview_bytes": 137520, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-young", "series": "variants", "run": "new-02-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.14, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "40ececc3e85c612ede7c98b29befc5ea336fe655e81f161ac3cecdb97760f9ab", "video": "/benchmarks/new-02-young.mp4", "poster": "/benchmarks/new-02-young.jpg", "api": "/benchmarks/new-02-young-api.json", "history": "/benchmarks/new-02-young-history.json", "source": "/benchmarks/new-02-young-source.json", "source_metadata": "/benchmarks/new-02-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 38.70967741935484, "mean_power_w": 174.37354838709678 }, "poster_seconds": 2, "notes": "The sampled frame retains the source\u2019s shadowed lighting and long hair. The mouth is open while parts of the eyes and cheeks remain dark, making fine facial detail harder to inspect.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 222694, "preview_video": "/benchmarks/new-02-young-preview.mp4", "preview_sha256": "42d7b671a53e21d27fa7e17c4c2dfb8e6a9fdf5e7bb3345a846888adee738b5a", "preview_bytes": 108328, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-02-avatar", "series": "variants", "run": "new-02-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.82, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "3a6a5de158b0b12af198e2bed84331b1873ed82b75114539253f84e8a239363b", "video": "/benchmarks/new-02-avatar.mp4", "poster": "/benchmarks/new-02-avatar.jpg", "api": "/benchmarks/new-02-avatar-api.json", "history": "/benchmarks/new-02-avatar-history.json", "source": "/benchmarks/new-02-avatar-source.json", "source_metadata": "/benchmarks/new-02-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 40.266666666666666, "mean_power_w": 178.40333333333334 }, "poster_seconds": 2, "notes": "The sampled frame retains the outdoor background and textured hair. Mouth opening and eye position differ from the source; inspect the jaw and neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 205359, "preview_video": "/benchmarks/new-02-avatar-preview.mp4", "preview_sha256": "746bcac91eb02c37d00b220b4bff556c3adbe3790ac667e99a16f5aafed0d635", "preview_bytes": 106130, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-lincoln", "series": "variants", "run": "new-03-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.692, "output_seconds": 4.0, "dimensions": [ 794, 1024 ], "sha256": "43f6ecab086b3759f6c76c4331e22230f43810637809c5b7f73812d12801cb3b", "video": "/benchmarks/new-03-lincoln.mp4", "poster": "/benchmarks/new-03-lincoln.jpg", "api": "/benchmarks/new-03-lincoln-api.json", "history": "/benchmarks/new-03-lincoln-history.json", "source": "/benchmarks/new-03-lincoln-source.json", "source_metadata": "/benchmarks/new-03-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4091.0, "mean_utilization_pct": 14.80952380952381, "mean_power_w": 92.75952380952381 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The archival oval/card border and seated body remain visible; the animated face occupies a small area in the preview.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 817020, "preview_video": "/benchmarks/new-03-lincoln-preview.mp4", "preview_sha256": "8fc4e906c4b0585e990a3805fe161f126a26573871421b8626d23429a1adc9bc", "preview_bytes": 49648, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-tubman", "series": "variants", "run": "new-03-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.735, "output_seconds": 4.0, "dimensions": [ 644, 1024 ], "sha256": "8ccb380d6f7a6473a2520931d6ff9c36f3a1ddd5bedc7796a3f6b47c9e5180a9", "video": "/benchmarks/new-03-tubman.mp4", "poster": "/benchmarks/new-03-tubman.jpg", "api": "/benchmarks/new-03-tubman-api.json", "history": "/benchmarks/new-03-tubman-history.json", "source": "/benchmarks/new-03-tubman-source.json", "source_metadata": "/benchmarks/new-03-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3963.0, "mean_utilization_pct": 9.904761904761905, "mean_power_w": 89.58142857142857 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The archival oval/card border and seated body remain visible; the animated face occupies a small area in the preview.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1242301, "preview_video": "/benchmarks/new-03-tubman-preview.mp4", "preview_sha256": "cdbd2ec3752c5969329b898e3f4b1f402ee92828bc624097255649b3bc1239f1", "preview_bytes": 53259, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-ai-woman", "series": "variants", "run": "new-03-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.918, "output_seconds": 4.0, "dimensions": [ 1254, 1254 ], "sha256": "9696911c002fc472989b8efe41f9fc35a737b59926c28ce1ec2c3e36e2bc6235", "video": "/benchmarks/new-03-ai-woman.mp4", "poster": "/benchmarks/new-03-ai-woman.jpg", "api": "/benchmarks/new-03-ai-woman-api.json", "history": "/benchmarks/new-03-ai-woman-history.json", "source": "/benchmarks/new-03-ai-woman-source.json", "source_metadata": "/benchmarks/new-03-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4027.0, "mean_utilization_pct": 6.854166666666667, "mean_power_w": 65.738125 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1157432, "preview_video": "/benchmarks/new-03-ai-woman-preview.mp4", "preview_sha256": "c0caa28bf029b2fc01a1e7c8ccda33271f76af1f10c11944e6990af8e26b8050", "preview_bytes": 79178, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-corporate", "series": "variants", "run": "new-03-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 39.155, "output_seconds": 4.0, "dimensions": [ 1116, 1034 ], "sha256": "accb987b8bd303097e0ec6def6d00e93ceaeeb96455be4d18aa754ab1793aacb", "video": "/benchmarks/new-03-corporate.mp4", "poster": "/benchmarks/new-03-corporate.jpg", "api": "/benchmarks/new-03-corporate-api.json", "history": "/benchmarks/new-03-corporate-history.json", "source": "/benchmarks/new-03-corporate-source.json", "source_metadata": "/benchmarks/new-03-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3963.0, "mean_utilization_pct": 8.657894736842104, "mean_power_w": 68.6471052631579 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 801436, "preview_video": "/benchmarks/new-03-corporate-preview.mp4", "preview_sha256": "bb77cbf8681c5df3919432cdefd157acfd7580a5a850aebf09feca143b985457", "preview_bytes": 69412, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-female", "series": "variants", "run": "new-03-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.129, "output_seconds": 4.0, "dimensions": [ 1114, 970 ], "sha256": "846de637faebe638e987f3a148d82a075ba78d906cfe5aaa077fab495a39dde7", "video": "/benchmarks/new-03-female.mp4", "poster": "/benchmarks/new-03-female.jpg", "api": "/benchmarks/new-03-female-api.json", "history": "/benchmarks/new-03-female-history.json", "source": "/benchmarks/new-03-female-source.json", "source_metadata": "/benchmarks/new-03-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3963.0, "mean_utilization_pct": 7.823529411764706, "mean_power_w": 75.58970588235294 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 838496, "preview_video": "/benchmarks/new-03-female-preview.mp4", "preview_sha256": "b20e614acf91e184d276dcbf8fa97a635f8fdab3ddc8078ee3695710e338b977", "preview_bytes": 66718, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-american", "series": "variants", "run": "new-03-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 33.207, "output_seconds": 4.0, "dimensions": [ 1120, 904 ], "sha256": "716d8887be33ca996d6552f92763344445c1c7f5ba3da89b116804a8af64726d", "video": "/benchmarks/new-03-american.mp4", "poster": "/benchmarks/new-03-american.jpg", "api": "/benchmarks/new-03-american-api.json", "history": "/benchmarks/new-03-american-history.json", "source": "/benchmarks/new-03-american-source.json", "source_metadata": "/benchmarks/new-03-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 9.375, "mean_power_w": 74.5628125 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 907584, "preview_video": "/benchmarks/new-03-american-preview.mp4", "preview_sha256": "ce318ff99773832810751c33a9179c41083251ea548bcf8beb10bea56ae0b4e4", "preview_bytes": 63278, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-young", "series": "variants", "run": "new-03-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 34.646, "output_seconds": 4.0, "dimensions": [ 1062, 856 ], "sha256": "c2b331c1c0790a965fa554eeed27b8841a4bcf09097208070899278e57914a67", "video": "/benchmarks/new-03-young.mp4", "poster": "/benchmarks/new-03-young.jpg", "api": "/benchmarks/new-03-young-api.json", "history": "/benchmarks/new-03-young-history.json", "source": "/benchmarks/new-03-young-source.json", "source_metadata": "/benchmarks/new-03-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3963.0, "mean_utilization_pct": 9.06060606060606, "mean_power_w": 79.80181818181818 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 642011, "preview_video": "/benchmarks/new-03-young-preview.mp4", "preview_sha256": "f8c6c90bd143a2f16dcbe65e68cae1af77ff8c3c3a3315ba5f169617f893709f", "preview_bytes": 62861, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-03-avatar", "series": "variants", "run": "new-03-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "full", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 36.731, "output_seconds": 4.0, "dimensions": [ 1000, 898 ], "sha256": "35de1b463795b6febc5fee750eddf18e10312da27a48fb8d58f807a0f0d4b170", "video": "/benchmarks/new-03-avatar.mp4", "poster": "/benchmarks/new-03-avatar.jpg", "api": "/benchmarks/new-03-avatar-api.json", "history": "/benchmarks/new-03-avatar-history.json", "source": "/benchmarks/new-03-avatar-source.json", "source_metadata": "/benchmarks/new-03-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3995.0, "mean_utilization_pct": 10.17142857142857, "mean_power_w": 79.43485714285714 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 759669, "preview_video": "/benchmarks/new-03-avatar-preview.mp4", "preview_sha256": "e4c370757215b1a262d8d2ffa10f83c8a8544eba818f2457e200643ac3c086b3", "preview_bytes": 71154, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-lincoln", "series": "variants", "run": "new-04-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 34.912, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "f9f664c7bd4e7a6df897491c90d96e4fda1cb4f5356ca0c254c096a25b0ca514", "video": "/benchmarks/new-04-lincoln.mp4", "poster": "/benchmarks/new-04-lincoln.jpg", "api": "/benchmarks/new-04-lincoln-api.json", "history": "/benchmarks/new-04-lincoln-history.json", "source": "/benchmarks/new-04-lincoln-source.json", "source_metadata": "/benchmarks/new-04-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 16.727272727272727, "mean_power_w": 93.83242424242424 }, "poster_seconds": 2, "notes": "The restored crop has smoother facial texture and more defined eye and beard edges than the unenhanced sample. This changes the appearance of the original scan; compare the raw result with the original likeness.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 215417, "preview_video": "/benchmarks/new-04-lincoln-preview.mp4", "preview_sha256": "744bd476bd7784bbc5468d2dd3fe84766ba2c86c0413fadc579c48039a50a2af", "preview_bytes": 94983, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-tubman", "series": "variants", "run": "new-04-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.034, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "8601a7bce51531065dc65a5d0888d673e81ca9b045ff904f5a4ab886a5c03d59", "video": "/benchmarks/new-04-tubman.mp4", "poster": "/benchmarks/new-04-tubman.jpg", "api": "/benchmarks/new-04-tubman-api.json", "history": "/benchmarks/new-04-tubman-history.json", "source": "/benchmarks/new-04-tubman-source.json", "source_metadata": "/benchmarks/new-04-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 16.764705882352942, "mean_power_w": 100.35617647058824 }, "poster_seconds": 2, "notes": "The restored crop has smoother skin and more defined eye edges than the unenhanced sample. The card border and seated body remain outside the crop; inspect the mouth and original likeness in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 235506, "preview_video": "/benchmarks/new-04-tubman-preview.mp4", "preview_sha256": "d60d1a7dad9477d9d4e87b53248a9c5e3e2a8638c0bedaad4d085faf01951273", "preview_bytes": 102580, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-ai-woman", "series": "variants", "run": "new-04-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.98, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "bdacad1fe766f72de2f9de8dd74cc14bd27b59c5e2159874d4ce40f4d1cd6f39", "video": "/benchmarks/new-04-ai-woman.mp4", "poster": "/benchmarks/new-04-ai-woman.jpg", "api": "/benchmarks/new-04-ai-woman-api.json", "history": "/benchmarks/new-04-ai-woman-history.json", "source": "/benchmarks/new-04-ai-woman-source.json", "source_metadata": "/benchmarks/new-04-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 18.6, "mean_power_w": 96.36228571428572 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 197270, "preview_video": "/benchmarks/new-04-ai-woman-preview.mp4", "preview_sha256": "c39b86e188a96078d16c237443ea598f7ae6f1bf5f428ec2cab1601578b558fc", "preview_bytes": 110159, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-corporate", "series": "variants", "run": "new-04-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.331, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "77ba83adb34791c1c4e567d3665051d43cb783f703254739e607ebb44d79a178", "video": "/benchmarks/new-04-corporate.mp4", "poster": "/benchmarks/new-04-corporate.jpg", "api": "/benchmarks/new-04-corporate-api.json", "history": "/benchmarks/new-04-corporate-history.json", "source": "/benchmarks/new-04-corporate-source.json", "source_metadata": "/benchmarks/new-04-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 16.058823529411764, "mean_power_w": 100.59470588235294 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 253249, "preview_video": "/benchmarks/new-04-corporate-preview.mp4", "preview_sha256": "b039bd486f73b41ca1a3debe30bfd8a6dbb3d988107c531c4b7b1a3a0ab4ce4b", "preview_bytes": 120527, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-female", "series": "variants", "run": "new-04-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.625, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "0785031997fa4f884775bfdc10b87d40d075b9d078771a428291f111e6632930", "video": "/benchmarks/new-04-female.mp4", "poster": "/benchmarks/new-04-female.jpg", "api": "/benchmarks/new-04-female-api.json", "history": "/benchmarks/new-04-female-history.json", "source": "/benchmarks/new-04-female-source.json", "source_metadata": "/benchmarks/new-04-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 17.97142857142857, "mean_power_w": 95.62371428571429 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth are visible in the sampled speech frame; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 222336, "preview_video": "/benchmarks/new-04-female-preview.mp4", "preview_sha256": "cbf59a138bb78db9b3b1a2d5c17fbcd1de7eeadd282f9cfc3be823c0df9c3673", "preview_bytes": 117737, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-american", "series": "variants", "run": "new-04-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.397, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "ab790ad1ffe6738e351ac8721056a80d144e0dbc3716b0a7af64bd8ba7f49b40", "video": "/benchmarks/new-04-american.mp4", "poster": "/benchmarks/new-04-american.jpg", "api": "/benchmarks/new-04-american-api.json", "history": "/benchmarks/new-04-american-history.json", "source": "/benchmarks/new-04-american-source.json", "source_metadata": "/benchmarks/new-04-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4091.0, "mean_utilization_pct": 17.470588235294116, "mean_power_w": 100.1814705882353 }, "poster_seconds": 2, "notes": "The sampled frame retains warm background lighting. Mouth shape and visible teeth differ from the closed-mouth source; compare facial detail at the intended display size.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 219718, "preview_video": "/benchmarks/new-04-american-preview.mp4", "preview_sha256": "73a8709c193357be9d3fcd2cf8941077d8e457da7f85f2fd2ee4582e99b8208b", "preview_bytes": 118213, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-young", "series": "variants", "run": "new-04-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.445, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "ae4abde58e92812ea73e6a1c098f04f1b8c11bb61577f23c22065e364f2943e6", "video": "/benchmarks/new-04-young.mp4", "poster": "/benchmarks/new-04-young.jpg", "api": "/benchmarks/new-04-young-api.json", "history": "/benchmarks/new-04-young-history.json", "source": "/benchmarks/new-04-young-source.json", "source_metadata": "/benchmarks/new-04-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 15.852941176470589, "mean_power_w": 100.03735294117647 }, "poster_seconds": 2, "notes": "The sampled frame retains the source\u2019s shadowed lighting and long hair. The mouth is open while parts of the eyes and cheeks remain dark, making fine facial detail harder to inspect.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 197027, "preview_video": "/benchmarks/new-04-young-preview.mp4", "preview_sha256": "ba65083aa3ea8eb74c6b4fa6287f00e27683ab388f5000b62376912408edc8ad", "preview_bytes": 104475, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-04-avatar", "series": "variants", "run": "new-04-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 35.73, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "59caee63743e3208920af1f7831e21af2007382f8296651ffaad427149fb6c3f", "video": "/benchmarks/new-04-avatar.mp4", "poster": "/benchmarks/new-04-avatar.jpg", "api": "/benchmarks/new-04-avatar-api.json", "history": "/benchmarks/new-04-avatar-history.json", "source": "/benchmarks/new-04-avatar-source.json", "source_metadata": "/benchmarks/new-04-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3995.0, "mean_utilization_pct": 19.264705882352942, "mean_power_w": 99.2420588235294 }, "poster_seconds": 2, "notes": "The sampled frame retains the outdoor background and textured hair. Mouth opening and eye position differ from the source; inspect the jaw and neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 213469, "preview_video": "/benchmarks/new-04-avatar-preview.mp4", "preview_sha256": "f46947b1d4861f6f61621939732192f843c9b44f85795f133036a4025e5180bc", "preview_bytes": 112196, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-lincoln", "series": "variants", "run": "new-05-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.131, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "2bed2419de73c5b4db710e6a157e00622c64df2f133727ad8e2a3211fb262f15", "video": "/benchmarks/new-05-lincoln.mp4", "poster": "/benchmarks/new-05-lincoln.jpg", "api": "/benchmarks/new-05-lincoln-api.json", "history": "/benchmarks/new-05-lincoln-history.json", "source": "/benchmarks/new-05-lincoln-source.json", "source_metadata": "/benchmarks/new-05-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8859.0, "mean_utilization_pct": 29.95744680851064, "mean_power_w": 153.04148936170213 }, "poster_seconds": 2, "notes": "The restored crop has smoother facial texture and more defined eye and beard edges than the unenhanced sample. This changes the appearance of the original scan; compare the raw result with the original likeness.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 788814, "preview_video": "/benchmarks/new-05-lincoln-preview.mp4", "preview_sha256": "a97a832332db81b8ba8848eec0a149fa666e3d6464c05a7ae01312035d62cc4a", "preview_bytes": 87590, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-tubman", "series": "variants", "run": "new-05-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.249, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "233af0a05784cc9f9e01d5541e78336eadf89cf841af959b553e8b213572d434", "video": "/benchmarks/new-05-tubman.mp4", "poster": "/benchmarks/new-05-tubman.jpg", "api": "/benchmarks/new-05-tubman-api.json", "history": "/benchmarks/new-05-tubman-history.json", "source": "/benchmarks/new-05-tubman-source.json", "source_metadata": "/benchmarks/new-05-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 31.82608695652174, "mean_power_w": 154.55 }, "poster_seconds": 2, "notes": "The restored crop has smoother skin and more defined eye edges than the unenhanced sample. The card border and seated body remain outside the crop; inspect the mouth and original likeness in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 872548, "preview_video": "/benchmarks/new-05-tubman-preview.mp4", "preview_sha256": "b3be1e21fed8b1a4eb1e87e3f5c6def601113c212100dd55dff1f15d1ac462c1", "preview_bytes": 111505, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-ai-woman", "series": "variants", "run": "new-05-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.221, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "d5164575da52bf745f858896afabb148ac627306066accb233e14ac5dbe9e1b5", "video": "/benchmarks/new-05-ai-woman.mp4", "poster": "/benchmarks/new-05-ai-woman.jpg", "api": "/benchmarks/new-05-ai-woman-api.json", "history": "/benchmarks/new-05-ai-woman-history.json", "source": "/benchmarks/new-05-ai-woman-source.json", "source_metadata": "/benchmarks/new-05-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8955.0, "mean_utilization_pct": 29.72340425531915, "mean_power_w": 151.29510638297873 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 602215, "preview_video": "/benchmarks/new-05-ai-woman-preview.mp4", "preview_sha256": "14241d4ccc74e86f7ffbbb5d998a4b2c1eec5348c63265a542cd08008f60ae2f", "preview_bytes": 104025, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-corporate", "series": "variants", "run": "new-05-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.87, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "2488b9bbdb274356a76f8f3ff37d821bd4dcfb5ce98b2dfd88f8561346ab50ed", "video": "/benchmarks/new-05-corporate.mp4", "poster": "/benchmarks/new-05-corporate.jpg", "api": "/benchmarks/new-05-corporate-api.json", "history": "/benchmarks/new-05-corporate-history.json", "source": "/benchmarks/new-05-corporate-source.json", "source_metadata": "/benchmarks/new-05-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 32.319148936170215, "mean_power_w": 154.73021276595745 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 949030, "preview_video": "/benchmarks/new-05-corporate-preview.mp4", "preview_sha256": "dd5d48fde269598c665a8b33ffff02fec920e62f262947bdb593b3e1cb90679b", "preview_bytes": 126179, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-female", "series": "variants", "run": "new-05-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.047, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "0f36e4e67e1d0e6aab344fb0ef4a9310db247b298ea9a30ab06cb263d037cfe9", "video": "/benchmarks/new-05-female.mp4", "poster": "/benchmarks/new-05-female.jpg", "api": "/benchmarks/new-05-female-api.json", "history": "/benchmarks/new-05-female-history.json", "source": "/benchmarks/new-05-female-source.json", "source_metadata": "/benchmarks/new-05-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 30.106382978723403, "mean_power_w": 152.34829787234042 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth are visible in the sampled speech frame; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 754405, "preview_video": "/benchmarks/new-05-female-preview.mp4", "preview_sha256": "ba64101fbd74579027ea1cd7750bfd668bc67f361d04aafe06d31957131e555e", "preview_bytes": 122523, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-american", "series": "variants", "run": "new-05-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.224, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "eb5de0d43e7e6433137930b2332b8e2c90e2777b8fb887ce5f6abdd08815f050", "video": "/benchmarks/new-05-american.mp4", "poster": "/benchmarks/new-05-american.jpg", "api": "/benchmarks/new-05-american-api.json", "history": "/benchmarks/new-05-american-history.json", "source": "/benchmarks/new-05-american-source.json", "source_metadata": "/benchmarks/new-05-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 30.148936170212767, "mean_power_w": 153.38404255319148 }, "poster_seconds": 2, "notes": "The sampled frame retains warm background lighting. Mouth shape and visible teeth differ from the closed-mouth source; compare facial detail at the intended display size.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 927249, "preview_video": "/benchmarks/new-05-american-preview.mp4", "preview_sha256": "0c3e5d0e0da5f8613e83cc50c997d03221c81074a27019abe879c5e53d650a26", "preview_bytes": 142906, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-young", "series": "variants", "run": "new-05-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.963, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "5ac59f100198bb93975f4cfff5544548ad0cd2b8fe4e52c4c4b4a494f36c385d", "video": "/benchmarks/new-05-young.mp4", "poster": "/benchmarks/new-05-young.jpg", "api": "/benchmarks/new-05-young-api.json", "history": "/benchmarks/new-05-young-history.json", "source": "/benchmarks/new-05-young-source.json", "source_metadata": "/benchmarks/new-05-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 29.70212765957447, "mean_power_w": 152.85638297872342 }, "poster_seconds": 2, "notes": "The sampled frame retains the source\u2019s shadowed lighting and long hair. The mouth is open while parts of the eyes and cheeks remain dark, making fine facial detail harder to inspect.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 687040, "preview_video": "/benchmarks/new-05-young-preview.mp4", "preview_sha256": "24c893b4d19cbecdfebde110491e76a39caf62cc7a93cb4e1614051e4a671de4", "preview_bytes": 110785, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-05-avatar", "series": "variants", "run": "new-05-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 49.244, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "78a19f19e948a128b3f8cfeaf8c165b374bd5d7af661153cdb29f2d84741809d", "video": "/benchmarks/new-05-avatar.mp4", "poster": "/benchmarks/new-05-avatar.jpg", "api": "/benchmarks/new-05-avatar-api.json", "history": "/benchmarks/new-05-avatar-history.json", "source": "/benchmarks/new-05-avatar-source.json", "source_metadata": "/benchmarks/new-05-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 29.70212765957447, "mean_power_w": 154.15382978723403 }, "poster_seconds": 2, "notes": "The sampled frame retains the outdoor background and textured hair. Mouth opening and eye position differ from the source; inspect the jaw and neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 828161, "preview_video": "/benchmarks/new-05-avatar-preview.mp4", "preview_sha256": "4993360021b0ec28fc98ef352d44d1719fdfbce8d906a19b70499269621ad0be", "preview_bytes": 133131, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-lincoln", "series": "variants", "run": "new-06-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.098, "output_seconds": 4.0, "dimensions": [ 256, 336 ], "sha256": "88242818202fa514d4ebd05c06489f72e0af3731aea80b511f05b9ba295d63aa", "video": "/benchmarks/new-06-lincoln.mp4", "poster": "/benchmarks/new-06-lincoln.jpg", "api": "/benchmarks/new-06-lincoln-api.json", "history": "/benchmarks/new-06-lincoln-history.json", "source": "/benchmarks/new-06-lincoln-source.json", "source_metadata": "/benchmarks/new-06-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 10.619047619047619, "mean_power_w": 83.18047619047618 }, "poster_seconds": 2, "notes": "Resize retains the full-body scan, but the sampled frame shows distortion in the surrounding oval/card and body compared with the original. Inspect these scene edges as well as the small face before using the result.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 136757, "preview_video": "/benchmarks/new-06-lincoln-preview.mp4", "preview_sha256": "2697c06a6ceae3421974a429edab1ad7c8e8f67e10c52720e6171d188b4c04e5", "preview_bytes": 72233, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-lincoln-small", "series": "variants", "run": "new-06-lincoln-small", "date": "2026-10-13", "image_type": "Low-resolution archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln-low-res.jpg", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "Lincoln archive thumbnail", "input_provenance": "Library of Congress 150 px thumbnail, unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "a3b770e0e9006fb858488888ebff8b28eb17ab4d0f8d334ff45d635636dde700", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.698, "output_seconds": 4.0, "dimensions": [ 256, 336 ], "sha256": "d7f819e9eae13987abe67ba4a0e369b83472a78707261c65264d55f72b343b91", "video": "/benchmarks/new-06-lincoln-small.mp4", "poster": "/benchmarks/new-06-lincoln-small.jpg", "api": "/benchmarks/new-06-lincoln-small-api.json", "history": "/benchmarks/new-06-lincoln-small-history.json", "source": "/benchmarks/new-06-lincoln-small-source.json", "source_metadata": "/benchmarks/new-06-lincoln-small-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.95, "mean_power_w": 93.0435 }, "poster_seconds": 2, "notes": "The tiny full-body thumbnail remains heavily blurred in the sampled frame. The oval outline and body shape are warped, and the face is too small to inspect mouth or eye detail. This is a difficult-input result, not a recommended restoration workflow.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 110431, "preview_video": "/benchmarks/new-06-lincoln-small-preview.mp4", "preview_sha256": "486eb8a4c984f32769b6e5342ae959c2db01d25ca25d0cf26bbc9f1d3b4a6c46", "preview_bytes": 65758, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-tubman", "series": "variants", "run": "new-06-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.117, "output_seconds": 4.0, "dimensions": [ 256, 416 ], "sha256": "bd46d0b2a561126d7da28f06e7c430d00e0c9b2375314ecf0f9cedadf70e735c", "video": "/benchmarks/new-06-tubman.mp4", "poster": "/benchmarks/new-06-tubman.jpg", "api": "/benchmarks/new-06-tubman-api.json", "history": "/benchmarks/new-06-tubman-history.json", "source": "/benchmarks/new-06-tubman-source.json", "source_metadata": "/benchmarks/new-06-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 15.238095238095237, "mean_power_w": 92.27904761904762 }, "poster_seconds": 2, "notes": "Resize retains the full-body scan, but the sampled frame shows distortion in the surrounding oval/card and body compared with the original. Inspect these scene edges as well as the small face before using the result.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 194923, "preview_video": "/benchmarks/new-06-tubman-preview.mp4", "preview_sha256": "e89fed119d1f6c5a739827bad73e14f2c60b96f841e89b0281a3a278c924d10d", "preview_bytes": 96845, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-ai-woman", "series": "variants", "run": "new-06-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.68, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "543d8fb73b42a050edbb027e5db21dffbfefabb569b0070dd3d3ca3b89ccbb1d", "video": "/benchmarks/new-06-ai-woman.mp4", "poster": "/benchmarks/new-06-ai-woman.jpg", "api": "/benchmarks/new-06-ai-woman-api.json", "history": "/benchmarks/new-06-ai-woman-history.json", "source": "/benchmarks/new-06-ai-woman-source.json", "source_metadata": "/benchmarks/new-06-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 9.380952380952381, "mean_power_w": 83.55333333333334 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 119134, "preview_video": "/benchmarks/new-06-ai-woman-preview.mp4", "preview_sha256": "abddf99004059a6a1753cbebe18aa67808ca251bd9e28fd0d13ea6c608409c91", "preview_bytes": 64561, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-corporate", "series": "variants", "run": "new-06-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.054, "output_seconds": 4.0, "dimensions": [ 256, 240 ], "sha256": "6e8224b0a09258b99a36c13571f57e9ba392ebfc88f9af65e2a9292edbd4911e", "video": "/benchmarks/new-06-corporate.mp4", "poster": "/benchmarks/new-06-corporate.jpg", "api": "/benchmarks/new-06-corporate-api.json", "history": "/benchmarks/new-06-corporate-history.json", "source": "/benchmarks/new-06-corporate-source.json", "source_metadata": "/benchmarks/new-06-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 10.9, "mean_power_w": 90.143 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 109958, "preview_video": "/benchmarks/new-06-corporate-preview.mp4", "preview_sha256": "20a377868d86c508c9810ac366feeb1b4075147e520693d99abb61977155dfaf", "preview_bytes": 58193, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-female", "series": "variants", "run": "new-06-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.898, "output_seconds": 4.0, "dimensions": [ 256, 224 ], "sha256": "38e0cda22359c496d3dc60f1c9f094e52e494c4b3dc53e1d2658f1968da83d48", "video": "/benchmarks/new-06-female.mp4", "poster": "/benchmarks/new-06-female.jpg", "api": "/benchmarks/new-06-female-api.json", "history": "/benchmarks/new-06-female-history.json", "source": "/benchmarks/new-06-female-source.json", "source_metadata": "/benchmarks/new-06-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 12.75, "mean_power_w": 82.011 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 90037, "preview_video": "/benchmarks/new-06-female-preview.mp4", "preview_sha256": "b9fcb7df48c5c62d3507c864c665154c8a1bf0618f7b99d928b51f1efe3c2d31", "preview_bytes": 50848, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-american", "series": "variants", "run": "new-06-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.994, "output_seconds": 4.0, "dimensions": [ 256, 208 ], "sha256": "3cc3452fa5f94020e6af54ace0ab002d3597d715407ee858f429f88ee5eacb79", "video": "/benchmarks/new-06-american.mp4", "poster": "/benchmarks/new-06-american.jpg", "api": "/benchmarks/new-06-american-api.json", "history": "/benchmarks/new-06-american-history.json", "source": "/benchmarks/new-06-american-source.json", "source_metadata": "/benchmarks/new-06-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.85, "mean_power_w": 91.95 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 110220, "preview_video": "/benchmarks/new-06-american-preview.mp4", "preview_sha256": "7917b9ac1ee06128c7a6d31e35c7fe4b94623332c0a3377a27a240ff196a5748", "preview_bytes": 59752, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-young", "series": "variants", "run": "new-06-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.127, "output_seconds": 4.0, "dimensions": [ 256, 208 ], "sha256": "5182b5c2f21ba494919f3cb8b5443e9ecac048989a0720de985bf0942eb27878", "video": "/benchmarks/new-06-young.mp4", "poster": "/benchmarks/new-06-young.jpg", "api": "/benchmarks/new-06-young-api.json", "history": "/benchmarks/new-06-young-history.json", "source": "/benchmarks/new-06-young-source.json", "source_metadata": "/benchmarks/new-06-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 9.333333333333334, "mean_power_w": 87.29904761904761 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 90979, "preview_video": "/benchmarks/new-06-young-preview.mp4", "preview_sha256": "cbd8d7906a9b2bb985b188249c8807ee121beccd5ded9afbeb80b63b72574831", "preview_bytes": 50419, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-06-avatar", "series": "variants", "run": "new-06-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "resize", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.82, "output_seconds": 4.0, "dimensions": [ 256, 240 ], "sha256": "b27097f35c73d05cd091dcdace04f07e641629b24319223842e98818ce1a448a", "video": "/benchmarks/new-06-avatar.mp4", "poster": "/benchmarks/new-06-avatar.jpg", "api": "/benchmarks/new-06-avatar-api.json", "history": "/benchmarks/new-06-avatar-history.json", "source": "/benchmarks/new-06-avatar-source.json", "source_metadata": "/benchmarks/new-06-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 11.3, "mean_power_w": 90.9155 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 94375, "preview_video": "/benchmarks/new-06-avatar-preview.mp4", "preview_sha256": "ecdc1fe3eb6acd2e1a2180db0ff550b10ba6de478d12e7a0bd0f2ac79e960653", "preview_bytes": 54671, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-lincoln", "series": "variants", "run": "new-07-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 60.552, "output_seconds": 4.0, "dimensions": [ 1600, 2048 ], "sha256": "a09802cec35e732e33d56ae1a27b6022dc68d4b633c9bc47c9734bc405f65670", "video": "/benchmarks/new-07-lincoln.mp4", "poster": "/benchmarks/new-07-lincoln.jpg", "api": "/benchmarks/new-07-lincoln-api.json", "history": "/benchmarks/new-07-lincoln-history.json", "source": "/benchmarks/new-07-lincoln-source.json", "source_metadata": "/benchmarks/new-07-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8667.0, "mean_utilization_pct": 27.92982456140351, "mean_power_w": 137.58105263157896 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The archival oval/card border and seated body remain visible; the animated face occupies a small area in the preview.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 540905, "preview_video": "/benchmarks/new-07-lincoln-preview.mp4", "preview_sha256": "ca60da04ea27970d8b6aa5872566d8ba9c841f182a74dd512d09bcc104d5ed54", "preview_bytes": 49477, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-tubman", "series": "variants", "run": "new-07-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 57.267, "output_seconds": 4.0, "dimensions": [ 1296, 2048 ], "sha256": "2de01e10812f167f62c8ba66f53ac607c58d6e84414c260a5035ada3e41f9748", "video": "/benchmarks/new-07-tubman.mp4", "poster": "/benchmarks/new-07-tubman.jpg", "api": "/benchmarks/new-07-tubman-api.json", "history": "/benchmarks/new-07-tubman-history.json", "source": "/benchmarks/new-07-tubman-source.json", "source_metadata": "/benchmarks/new-07-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 28.925925925925927, "mean_power_w": 139.0048148148148 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The archival oval/card border and seated body remain visible; the animated face occupies a small area in the preview.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 691438, "preview_video": "/benchmarks/new-07-tubman-preview.mp4", "preview_sha256": "becd1158e7aabb0d675519e1c84dd41bb679068eaaaaf722fbcf283fbe759bd7", "preview_bytes": 52951, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-ai-woman", "series": "variants", "run": "new-07-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 103.895, "output_seconds": 4.0, "dimensions": [ 2512, 2512 ], "sha256": "943f5a497c1eb237f847854ff4aa0d0545456a4956d3dd566f71ca90ad15f921", "video": "/benchmarks/new-07-ai-woman.mp4", "poster": "/benchmarks/new-07-ai-woman.jpg", "api": "/benchmarks/new-07-ai-woman-api.json", "history": "/benchmarks/new-07-ai-woman-history.json", "source": "/benchmarks/new-07-ai-woman-source.json", "source_metadata": "/benchmarks/new-07-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8955.0, "mean_utilization_pct": 15.515151515151516, "mean_power_w": 104.89909090909092 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 2594801, "preview_video": "/benchmarks/new-07-ai-woman-preview.mp4", "preview_sha256": "558f120e9bd2997ef41c47d9a0c2dcdae0fa1f530e0a0691bb1a7ea450d86bd2", "preview_bytes": 79808, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-corporate", "series": "variants", "run": "new-07-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 83.815, "output_seconds": 4.0, "dimensions": [ 2240, 2080 ], "sha256": "1241cdf3198e4d159870a7cd2b160d10b97a6c12252c4735f27a918d5b3b81be", "video": "/benchmarks/new-07-corporate.mp4", "poster": "/benchmarks/new-07-corporate.jpg", "api": "/benchmarks/new-07-corporate-api.json", "history": "/benchmarks/new-07-corporate-history.json", "source": "/benchmarks/new-07-corporate-source.json", "source_metadata": "/benchmarks/new-07-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 19.0126582278481, "mean_power_w": 117.32683544303798 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 2006610, "preview_video": "/benchmarks/new-07-corporate-preview.mp4", "preview_sha256": "42a9cb337f75d69d69731142223ed7c1e741711a9e545a87e4e5ef8662846ed4", "preview_bytes": 71942, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-female", "series": "variants", "run": "new-07-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 78.188, "output_seconds": 4.0, "dimensions": [ 2240, 1952 ], "sha256": "58e8b48fcf195f504bc995066c44f6ecedacc0154362cdf61bf96228e14bf512", "video": "/benchmarks/new-07-female.mp4", "poster": "/benchmarks/new-07-female.jpg", "api": "/benchmarks/new-07-female-api.json", "history": "/benchmarks/new-07-female-history.json", "source": "/benchmarks/new-07-female-source.json", "source_metadata": "/benchmarks/new-07-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 20.027027027027028, "mean_power_w": 121.31256756756757 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1885131, "preview_video": "/benchmarks/new-07-female-preview.mp4", "preview_sha256": "07f4e5e5c8580428677cb6a0eab3c87b397b5da777b3a70ca79cca39708fecc5", "preview_bytes": 73412, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-american", "series": "variants", "run": "new-07-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 76.462, "output_seconds": 4.0, "dimensions": [ 2240, 1808 ], "sha256": "5177a4abf708e51ad43a9ebe0afd967b97acc2a5d03fca738a5c68aef60338ca", "video": "/benchmarks/new-07-american.mp4", "poster": "/benchmarks/new-07-american.jpg", "api": "/benchmarks/new-07-american-api.json", "history": "/benchmarks/new-07-american-history.json", "source": "/benchmarks/new-07-american-source.json", "source_metadata": "/benchmarks/new-07-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 22.01388888888889, "mean_power_w": 122.29166666666667 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1678231, "preview_video": "/benchmarks/new-07-american-preview.mp4", "preview_sha256": "f23836cf60a8f5e66197650b4267fcf0617bc725bf21f0388d12c7678bc33602", "preview_bytes": 66954, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-young", "series": "variants", "run": "new-07-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 75.611, "output_seconds": 4.0, "dimensions": [ 2128, 1712 ], "sha256": "a9420bf11d54e023800a6a4c2c5db291f217604825a1f2154e751ce810eed66c", "video": "/benchmarks/new-07-young.mp4", "poster": "/benchmarks/new-07-young.jpg", "api": "/benchmarks/new-07-young-api.json", "history": "/benchmarks/new-07-young-history.json", "source": "/benchmarks/new-07-young-source.json", "source_metadata": "/benchmarks/new-07-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 22.875, "mean_power_w": 124.01347222222222 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1493500, "preview_video": "/benchmarks/new-07-young-preview.mp4", "preview_sha256": "bbf3b16dedf858045050e4dbe3bddd7c1f6dc2dbb2a8c954083b710f9ac8515f", "preview_bytes": 68870, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-07-avatar", "series": "variants", "run": "new-07-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "full", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 79.381, "output_seconds": 4.0, "dimensions": [ 2000, 1808 ], "sha256": "7d6022a290564d45dc8c96f539b8439a83b87264bb9356edcc4fbbae9ead90af", "video": "/benchmarks/new-07-avatar.mp4", "poster": "/benchmarks/new-07-avatar.jpg", "api": "/benchmarks/new-07-avatar-api.json", "history": "/benchmarks/new-07-avatar-history.json", "source": "/benchmarks/new-07-avatar-source.json", "source_metadata": "/benchmarks/new-07-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 21.18421052631579, "mean_power_w": 121.19276315789473 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1946790, "preview_video": "/benchmarks/new-07-avatar-preview.mp4", "preview_sha256": "b919cbb7a70bd0e821a5a93fceb7272852dc4ad0b8baa2cfbe0108002ffa9f3c", "preview_bytes": 77455, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-lincoln", "series": "variants", "run": "new-08-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 50.644, "output_seconds": 4.0, "dimensions": [ 1024, 1344 ], "sha256": "d51c1f8fff5d5a5b3a42ec1c59d86b43e5285d1763b201fde6902c3ea5310fd5", "video": "/benchmarks/new-08-lincoln.mp4", "poster": "/benchmarks/new-08-lincoln.jpg", "api": "/benchmarks/new-08-lincoln-api.json", "history": "/benchmarks/new-08-lincoln-history.json", "source": "/benchmarks/new-08-lincoln-source.json", "source_metadata": "/benchmarks/new-08-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 30.020833333333332, "mean_power_w": 150.27875 }, "poster_seconds": 2, "notes": "Resize retains the full-body scan, but the sampled frame shows distortion in the surrounding oval/card and body compared with the original. Inspect these scene edges as well as the small face before using the result.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 555951, "preview_video": "/benchmarks/new-08-lincoln-preview.mp4", "preview_sha256": "bf0af20878dc45010e9be26a9e1fad08fd4ac32462d53cb0549deb34ea5696f3", "preview_bytes": 72820, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-lincoln-small", "series": "variants", "run": "new-08-lincoln-small", "date": "2026-10-13", "image_type": "Low-resolution archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln-low-res.jpg", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "Lincoln archive thumbnail", "input_provenance": "Library of Congress 150 px thumbnail, unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "a3b770e0e9006fb858488888ebff8b28eb17ab4d0f8d334ff45d635636dde700", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 50.539, "output_seconds": 4.0, "dimensions": [ 1024, 1344 ], "sha256": "4ea2b1ae4b685b2a086162d40ba210317515c116598781bacffe2098bfa42c8a", "video": "/benchmarks/new-08-lincoln-small.mp4", "poster": "/benchmarks/new-08-lincoln-small.jpg", "api": "/benchmarks/new-08-lincoln-small-api.json", "history": "/benchmarks/new-08-lincoln-small-history.json", "source": "/benchmarks/new-08-lincoln-small-source.json", "source_metadata": "/benchmarks/new-08-lincoln-small-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 29.625, "mean_power_w": 150.34791666666666 }, "poster_seconds": 2, "notes": "The tiny full-body thumbnail remains heavily blurred in the sampled frame. The oval outline and body shape are warped, and the face is too small to inspect mouth or eye detail. This is a difficult-input result, not a recommended restoration workflow.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 590057, "preview_video": "/benchmarks/new-08-lincoln-small-preview.mp4", "preview_sha256": "246c50bfeafe2805346584cb6f05e476a0241a499623c61c1b6518c414ce5e5b", "preview_bytes": 78067, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-tubman", "series": "variants", "run": "new-08-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 51.163, "output_seconds": 4.0, "dimensions": [ 1024, 1632 ], "sha256": "ac7648c4fbb4f832617c59d02d1324dcbabcc7b6cd84b42e31f2d46754b01a47", "video": "/benchmarks/new-08-tubman.mp4", "poster": "/benchmarks/new-08-tubman.jpg", "api": "/benchmarks/new-08-tubman-api.json", "history": "/benchmarks/new-08-tubman-history.json", "source": "/benchmarks/new-08-tubman-source.json", "source_metadata": "/benchmarks/new-08-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 29.244897959183675, "mean_power_w": 146.72408163265305 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The archival oval/card border and seated body remain visible; the animated face occupies a small area in the preview.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 1134100, "preview_video": "/benchmarks/new-08-tubman-preview.mp4", "preview_sha256": "0ee102549305c4c2a2cd3986c18e1c1c78a36798e7543bc03d3e0bc315874485", "preview_bytes": 102492, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-ai-woman", "series": "variants", "run": "new-08-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.903, "output_seconds": 4.0, "dimensions": [ 1024, 1024 ], "sha256": "4ee8ad63271714f3f36dd7e00e8ba037cc03e558f23aa910e8a5826c8662f239", "video": "/benchmarks/new-08-ai-woman.mp4", "poster": "/benchmarks/new-08-ai-woman.jpg", "api": "/benchmarks/new-08-ai-woman-api.json", "history": "/benchmarks/new-08-ai-woman-history.json", "source": "/benchmarks/new-08-ai-woman-source.json", "source_metadata": "/benchmarks/new-08-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 30.659574468085108, "mean_power_w": 151.63872340425533 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 501021, "preview_video": "/benchmarks/new-08-ai-woman-preview.mp4", "preview_sha256": "28f862d3ca477f1219ea3b9088ac1eaff2bc99aad4f6f41c07a7dd6ac77fd8a6", "preview_bytes": 95554, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-corporate", "series": "variants", "run": "new-08-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.974, "output_seconds": 4.0, "dimensions": [ 1024, 960 ], "sha256": "d69fbfbe1dea9401fcbda695caa703844c4cae0e01f2658034ed4a4b437c3b78", "video": "/benchmarks/new-08-corporate.mp4", "poster": "/benchmarks/new-08-corporate.jpg", "api": "/benchmarks/new-08-corporate-api.json", "history": "/benchmarks/new-08-corporate-history.json", "source": "/benchmarks/new-08-corporate-source.json", "source_metadata": "/benchmarks/new-08-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 30.32608695652174, "mean_power_w": 151.11413043478262 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 588794, "preview_video": "/benchmarks/new-08-corporate-preview.mp4", "preview_sha256": "4225c1780c803c543c1bec97cf4c802cc67c4769f4d5a026199a22df3d20c2da", "preview_bytes": 90403, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-female", "series": "variants", "run": "new-08-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.277, "output_seconds": 4.0, "dimensions": [ 1024, 896 ], "sha256": "fdda417850aaa043231249562c1e8646b0f620f422ee35f3e4dc839453c3fb1c", "video": "/benchmarks/new-08-female.mp4", "poster": "/benchmarks/new-08-female.jpg", "api": "/benchmarks/new-08-female-api.json", "history": "/benchmarks/new-08-female-history.json", "source": "/benchmarks/new-08-female-source.json", "source_metadata": "/benchmarks/new-08-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 33.52173913043478, "mean_power_w": 151.38521739130434 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 454892, "preview_video": "/benchmarks/new-08-female-preview.mp4", "preview_sha256": "4f7b15cbe73fb9eb19385fd771994e4659abfa25f34771caaacef33dd8ce894d", "preview_bytes": 86086, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-american", "series": "variants", "run": "new-08-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.229, "output_seconds": 4.0, "dimensions": [ 1024, 832 ], "sha256": "a49b77377cb2c83acc14a7b470f2a9554abcbd23865b63854e93f24fb00c7fed", "video": "/benchmarks/new-08-american.mp4", "poster": "/benchmarks/new-08-american.jpg", "api": "/benchmarks/new-08-american-api.json", "history": "/benchmarks/new-08-american-history.json", "source": "/benchmarks/new-08-american-source.json", "source_metadata": "/benchmarks/new-08-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 32.19565217391305, "mean_power_w": 154.44521739130434 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 545397, "preview_video": "/benchmarks/new-08-american-preview.mp4", "preview_sha256": "c9f65294ff1cdd5507765e876962d8874796f1fe6e51a07c60d5ff0cc031b7f7", "preview_bytes": 96624, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-young", "series": "variants", "run": "new-08-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.167, "output_seconds": 4.0, "dimensions": [ 1024, 832 ], "sha256": "fbef46d4d4b4d7c1a487d58ad6cbfdd6b1ba288440b65bfa207493f3b695b999", "video": "/benchmarks/new-08-young.mp4", "poster": "/benchmarks/new-08-young.jpg", "api": "/benchmarks/new-08-young-api.json", "history": "/benchmarks/new-08-young-history.json", "source": "/benchmarks/new-08-young-source.json", "source_metadata": "/benchmarks/new-08-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 32.32608695652174, "mean_power_w": 153.91543478260868 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 471414, "preview_video": "/benchmarks/new-08-young-preview.mp4", "preview_sha256": "5aedb56aa2618378ce584d7d7ddf690629aa4a6a45b8c83cb41a79433dc03f23", "preview_bytes": 86867, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-08-avatar", "series": "variants", "run": "new-08-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "resize", "still": true, "enhancer": "gfpgan", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 48.459, "output_seconds": 4.0, "dimensions": [ 1024, 928 ], "sha256": "06a5036ebb3c8c92270c88c0f3a0e183aa956a8684d39b48281d2f5469881ba1", "video": "/benchmarks/new-08-avatar.mp4", "poster": "/benchmarks/new-08-avatar.jpg", "api": "/benchmarks/new-08-avatar-api.json", "history": "/benchmarks/new-08-avatar-history.json", "source": "/benchmarks/new-08-avatar-source.json", "source_metadata": "/benchmarks/new-08-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8635.0, "mean_utilization_pct": 33.65217391304348, "mean_power_w": 150.40326086956523 }, "poster_seconds": 2, "notes": "The sampled frame retains the wider input scene. Resize renders this wider framing directly; inspect background and body edges as well as the face. The presenter and surrounding background remain in the composition; inspect the face-to-neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 544006, "preview_video": "/benchmarks/new-08-avatar-preview.mp4", "preview_sha256": "e607cc7706b7f7d136191d5cd517aca331dc151c2ec7b688bf0fcb3a22acdb01", "preview_bytes": 96703, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-lincoln", "series": "variants", "run": "new-09-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.79, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "32d6474193e952a2152f1af4fe13853212ffd251458f08add86d27c47f091752", "video": "/benchmarks/new-09-lincoln.mp4", "poster": "/benchmarks/new-09-lincoln.jpg", "api": "/benchmarks/new-09-lincoln-api.json", "history": "/benchmarks/new-09-lincoln-history.json", "source": "/benchmarks/new-09-lincoln-source.json", "source_metadata": "/benchmarks/new-09-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.65, "mean_power_w": 89.772 }, "poster_seconds": 2, "notes": "The sampled face crop retains the scan\u2019s sepia tone and soft texture. Teeth and the open-mouth edge appear less uniform than the closed mouth in the original scan.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 68600, "preview_video": "/benchmarks/new-09-lincoln-preview.mp4", "preview_sha256": "87eb80919b7d552ed48141724d855881abc212eb8373c8a90b0343049ff30a67", "preview_bytes": 44912, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-tubman", "series": "variants", "run": "new-09-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 20.718, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "c936286e5696509a1f29d18b3e97a4b4b6b97ef1e3935e825ef7eb00eb61de29", "video": "/benchmarks/new-09-tubman.mp4", "poster": "/benchmarks/new-09-tubman.jpg", "api": "/benchmarks/new-09-tubman-api.json", "history": "/benchmarks/new-09-tubman-history.json", "source": "/benchmarks/new-09-tubman-source.json", "source_metadata": "/benchmarks/new-09-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 15.25, "mean_power_w": 79.01350000000001 }, "poster_seconds": 2, "notes": "The sampled crop enlarges the archival face and shows an open mouth. Facial edges remain soft; the original card border and seated body are outside this crop.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 76342, "preview_video": "/benchmarks/new-09-tubman-preview.mp4", "preview_sha256": "405ee182a867e75463b92283bf4441af6fc3415efba1180628647f5b70a66039", "preview_bytes": 48787, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-ai-woman", "series": "variants", "run": "new-09-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.466, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "29e48fdc6c6b3561f1538e78c14af26765327370003042b75efdbf11544e0558", "video": "/benchmarks/new-09-ai-woman.mp4", "poster": "/benchmarks/new-09-ai-woman.jpg", "api": "/benchmarks/new-09-ai-woman-api.json", "history": "/benchmarks/new-09-ai-woman-history.json", "source": "/benchmarks/new-09-ai-woman-source.json", "source_metadata": "/benchmarks/new-09-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 14.142857142857142, "mean_power_w": 89.48619047619047 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 95691, "preview_video": "/benchmarks/new-09-ai-woman-preview.mp4", "preview_sha256": "3bb75b4b140bba6c9e1098ca8b92cf5f56b13c000b5cd9781322e3dfe4fcabb3", "preview_bytes": 55232, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-corporate", "series": "variants", "run": "new-09-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.221, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "4cf22980dc73223d384589f4513d1ff0d73a5531d9c4ad5fe266fb086abf248c", "video": "/benchmarks/new-09-corporate.mp4", "poster": "/benchmarks/new-09-corporate.jpg", "api": "/benchmarks/new-09-corporate-api.json", "history": "/benchmarks/new-09-corporate-history.json", "source": "/benchmarks/new-09-corporate-source.json", "source_metadata": "/benchmarks/new-09-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 16.571428571428573, "mean_power_w": 82.36 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 120025, "preview_video": "/benchmarks/new-09-corporate-preview.mp4", "preview_sha256": "46a8a66ba9fc8ffd5ba79a682d2457900111507d9bf13eb2a6189eab4dd63cf3", "preview_bytes": 61008, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-female", "series": "variants", "run": "new-09-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.108, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "fd8aa619f7b8a4c8a72ac4d93fcccc9f4fe36367e82b2f7f8585ed7f86816758", "video": "/benchmarks/new-09-female.mp4", "poster": "/benchmarks/new-09-female.jpg", "api": "/benchmarks/new-09-female-api.json", "history": "/benchmarks/new-09-female-history.json", "source": "/benchmarks/new-09-female-source.json", "source_metadata": "/benchmarks/new-09-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 16.761904761904763, "mean_power_w": 91.05095238095238 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth are visible in the sampled speech frame; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 102426, "preview_video": "/benchmarks/new-09-female-preview.mp4", "preview_sha256": "226fb50333a96a2c1435ec32aa513b5bff0bf9b39a30c1d7a831e3249237fe6b", "preview_bytes": 56050, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-american", "series": "variants", "run": "new-09-american", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-american-female-source.png", "audio": "/benchmarks/inputs/narration-1-4s.wav", "input_title": "American narrator portrait", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Female instruction", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "eae7a5d17c8c3967bbebf491627d4f3b40642badf869cbb2cf1772c4dccc92b7", "audio": "9276677c78a848552e1fc0eeee36c9330c8eddee88e836af074c02c2a6739a6b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.086, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "3522d85188450c81882b4143650818eb003b6723a7e252f29017ae12962d891e", "video": "/benchmarks/new-09-american.mp4", "poster": "/benchmarks/new-09-american.jpg", "api": "/benchmarks/new-09-american-api.json", "history": "/benchmarks/new-09-american-history.json", "source": "/benchmarks/new-09-american-source.json", "source_metadata": "/benchmarks/new-09-american-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4091.0, "mean_utilization_pct": 12.333333333333334, "mean_power_w": 87.97761904761904 }, "poster_seconds": 2, "notes": "The sampled frame retains warm background lighting. Mouth shape and visible teeth differ from the closed-mouth source; compare facial detail at the intended display size.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 104873, "preview_video": "/benchmarks/new-09-american-preview.mp4", "preview_sha256": "24dc341148134a6ad050d12b6dace9f3f99923a9087878f0d6dcf3bcf0ff6000", "preview_bytes": 56051, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-young", "series": "variants", "run": "new-09-young", "date": "2026-10-13", "image_type": "Casual portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-young-creator-source.png", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Young creator", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "9b50790feca88c610cf641220df86741b4735c3a885176b96d4a6aa5e8cfefd9", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.031, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "f8310ccaad06eede2bb2132559f67c1e072c87a2e4f0b42e8cf4c17b1285dccd", "video": "/benchmarks/new-09-young.mp4", "poster": "/benchmarks/new-09-young.jpg", "api": "/benchmarks/new-09-young-api.json", "history": "/benchmarks/new-09-young-history.json", "source": "/benchmarks/new-09-young-source.json", "source_metadata": "/benchmarks/new-09-young-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 4059.0, "mean_utilization_pct": 9.666666666666666, "mean_power_w": 87.85809523809525 }, "poster_seconds": 2, "notes": "The sampled frame retains the source\u2019s shadowed lighting and long hair. The mouth is open while parts of the eyes and cheeks remain dark, making fine facial detail harder to inspect.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 95319, "preview_video": "/benchmarks/new-09-young-preview.mp4", "preview_sha256": "b2761188972022332306764fb0c99c505336fa8b7e8ed5b28a8421fdfd66494b", "preview_bytes": 52585, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-09-avatar", "series": "variants", "run": "new-09-avatar", "date": "2026-10-13", "image_type": "Avatar portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-male-avatar-source.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Male avatar", "input_provenance": "Existing website avatar example; unchanged.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "36243a55cd14ad0aa287295a7cdfcad1a2d26a4044009417f929aaa527fde244", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 256, "preprocess": "crop", "still": true, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 21.013, "output_seconds": 4.0, "dimensions": [ 256, 256 ], "sha256": "0af82df8079d6727fe72b09a6bac4d1e92dd08963b76f2fb677c7c1f44541043", "video": "/benchmarks/new-09-avatar.mp4", "poster": "/benchmarks/new-09-avatar.jpg", "api": "/benchmarks/new-09-avatar-api.json", "history": "/benchmarks/new-09-avatar-history.json", "source": "/benchmarks/new-09-avatar-source.json", "source_metadata": "/benchmarks/new-09-avatar-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 3995.0, "mean_utilization_pct": 14.4, "mean_power_w": 88.7025 }, "poster_seconds": 2, "notes": "The sampled frame retains the outdoor background and textured hair. Mouth opening and eye position differ from the source; inspect the jaw and neck boundary in the raw output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 93071, "preview_video": "/benchmarks/new-09-avatar-preview.mp4", "preview_sha256": "a380b806cbaf5c82a68d4494ab84cf22b983b75d6ea7d5c5dc19b7a4a92ec241", "preview_bytes": 54637, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-10-lincoln", "series": "variants", "run": "new-10-lincoln", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/lincoln.jpg", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Abraham Lincoln, 1863", "input_provenance": "Library of Congress, Alexander Gardner, 1863; original scan, unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b60ef17bb0ad3de56dc8231fcf37fdb78a38cd4dece8f9dc5ac6b2d889fe0e9f", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.609, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "86474c9df7a4da26e267d95ed50d1738759ee97040868af797baa63831b53848", "video": "/benchmarks/new-10-lincoln.mp4", "poster": "/benchmarks/new-10-lincoln.jpg", "api": "/benchmarks/new-10-lincoln-api.json", "history": "/benchmarks/new-10-lincoln-history.json", "source": "/benchmarks/new-10-lincoln-source.json", "source_metadata": "/benchmarks/new-10-lincoln-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8859.0, "mean_utilization_pct": 36.733333333333334, "mean_power_w": 171.25666666666666 }, "poster_seconds": 2, "notes": "The sampled face crop retains the scan\u2019s sepia tone and soft texture. Teeth and the open-mouth edge appear less uniform than the closed mouth in the original scan.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 151831, "preview_video": "/benchmarks/new-10-lincoln-preview.mp4", "preview_sha256": "74b7c78455b56be244dd60d29c479df6851e177d638d0ded287fa54c7048488b", "preview_bytes": 96905, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-10-tubman", "series": "variants", "run": "new-10-tubman", "date": "2026-10-13", "image_type": "Archival photo", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/tubman.jpg", "audio": "/benchmarks/inputs/narration-2-4s.wav", "input_title": "Harriet Tubman, 1868 or 1869", "input_provenance": "Library of Congress, Benjamin F. Powelson, 1868 or 1869; original scan, unchanged.", "audio_style": "American conversational narration", "clip_description": "4-second excerpt of existing website narration. Synthetic speech animation paired with a website narration excerpt; not a historical voice recording.", "input_hashes": { "image": "b49c2f1e1d45c9441f4c505c8bbd793520de9decc86670b12491120498f15fe5", "audio": "01f86f92002a3f7aeb7bf03dc63b39db1b624189410f0c1f117748a775538e2b" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 30.527, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "7f32be5fc512363cab0dba126590155778a9e65e7d117d340aa33267a61efc01", "video": "/benchmarks/new-10-tubman.mp4", "poster": "/benchmarks/new-10-tubman.jpg", "api": "/benchmarks/new-10-tubman-api.json", "history": "/benchmarks/new-10-tubman-history.json", "source": "/benchmarks/new-10-tubman-source.json", "source_metadata": "/benchmarks/new-10-tubman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 38.41379310344828, "mean_power_w": 177.18275862068967 }, "poster_seconds": 2, "notes": "The sampled crop enlarges the archival face and shows an open mouth. Facial edges remain soft; the original card border and seated body are outside this crop.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 165999, "preview_video": "/benchmarks/new-10-tubman-preview.mp4", "preview_sha256": "c4a7f6fe09ad6dfbee7dd92b0d2ac4f85e32620e8c19b4e00cd5389076c8672c", "preview_bytes": 101666, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-10-ai-woman", "series": "variants", "run": "new-10-ai-woman", "date": "2026-10-13", "image_type": "AI-generated portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/ai-woman.png", "audio": "/benchmarks/inputs/narration-3-4s.wav", "input_title": "Fictional AI presenter", "input_provenance": "Generated for this benchmark with the built-in image generation tool on 13 October 2026.", "audio_style": "Young-creator narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "7aa6e7cff8d5f3d092d0f2486a35ab9ae6cb35070f17d62b39a1eb94ceb877d2", "audio": "1b7157c91808c6b6daaae613e6282e42ad09dffb31c0f2a4173ac8c01c9f9e88" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.678, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "14112fcc5db71c83a4dc15d5a8643ffafb9b0a9f6f69ba7def099c28c6f35308", "video": "/benchmarks/new-10-ai-woman.mp4", "poster": "/benchmarks/new-10-ai-woman.jpg", "api": "/benchmarks/new-10-ai-woman-api.json", "history": "/benchmarks/new-10-ai-woman-history.json", "source": "/benchmarks/new-10-ai-woman-source.json", "source_metadata": "/benchmarks/new-10-ai-woman-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8955.0, "mean_utilization_pct": 40.2, "mean_power_w": 172.78033333333335 }, "poster_seconds": 2, "notes": "The sampled frame shows the fictional presenter with visible teeth and a plain gray background. Compare the mouth and tooth detail with the input; a sharp still frame does not establish lip-sync accuracy.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 260361, "preview_video": "/benchmarks/new-10-ai-woman-preview.mp4", "preview_sha256": "7a2996f15707782910ef6aa47d7aa719c61cb19b236ea8db838fe4129f1fda54", "preview_bytes": 128174, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-10-corporate", "series": "variants", "run": "new-10-corporate", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-corporate-presenter-source.png", "audio": "/benchmarks/inputs/narration-4-4s.wav", "input_title": "Corporate presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Corporate narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "834fef997641c07483b70053ac6e9e47fa3bdd7a91c8f44596fa802309d5af6a", "audio": "681359d088f6bedf363f2c2161b6474ad64a61c479cbef24c4be8e0af8a717b3" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.447, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "5ad340e229d390c045ec718ef7dd4a318641c8996dc2d859da20aaf520caee2a", "video": "/benchmarks/new-10-corporate.mp4", "poster": "/benchmarks/new-10-corporate.jpg", "api": "/benchmarks/new-10-corporate-api.json", "history": "/benchmarks/new-10-corporate-history.json", "source": "/benchmarks/new-10-corporate-source.json", "source_metadata": "/benchmarks/new-10-corporate-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 36.733333333333334, "mean_power_w": 175.524 }, "poster_seconds": 2, "notes": "The sampled frame retains the dark studio background and facial hair. The mouth and eye opening vary between runs, and beard detail appears soft in the smaller base renders.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 446121, "preview_video": "/benchmarks/new-10-corporate-preview.mp4", "preview_sha256": "2aa4a2dc1c2c562935ab90c7c3fb8aa2e57ba7f6b39e8c581c8e9f6f4f7a1f94", "preview_bytes": 169779, "notes_review": "Reviewed the 2-second output frame against the original input." }, { "id": "new-10-female", "series": "variants", "run": "new-10-female", "date": "2026-10-13", "image_type": "Studio portrait", "language": "English", "gpu": "RTX 4090", "image": "/benchmarks/inputs/sadtalker-female-portrait-source.png", "audio": "/benchmarks/inputs/narration-0-4s.wav", "input_title": "Female presenter", "input_provenance": "Existing website demo source; unchanged.", "audio_style": "Male narration", "clip_description": "4-second excerpt of existing website narration; no new speech synthesized for this run.", "input_hashes": { "image": "e14555c32823208406fd08591addb224c3eaccb2bc9894821cf5bf7303d4b405", "audio": "9180312db9fbbf21800ed3be376cd9e2818cc2b7cc5735cb3a5d654b5937c451" }, "settings": { "size": 512, "preprocess": "crop", "still": false, "enhancer": "none", "batch_size": 2, "expression_scale": 1.0, "pose_style": 0 }, "execution_seconds": 31.336, "output_seconds": 4.0, "dimensions": [ 512, 512 ], "sha256": "f4d6553fb2ce427946e37b7fc2540d9d1f571482c3c7f0dafc819a4ae8a3e59a", "video": "/benchmarks/new-10-female.mp4", "poster": "/benchmarks/new-10-female.jpg", "api": "/benchmarks/new-10-female-api.json", "history": "/benchmarks/new-10-female-history.json", "source": "/benchmarks/new-10-female-source.json", "source_metadata": "/benchmarks/new-10-female-source.json", "environment": { "gpu": "NVIDIA GeForce RTX 4090, 24564 MiB, 570.86.16", "comfyui_revision": "73c9bad4d21e7addbe1d13bc92eee0f1431b017d", "sadtalker_revision": "cd4c0465ae0b54a6f85af57f5c65fec9fe23e7f8", "worker": { "python": "3.10.21", "torch": "2.1.2+cu121", "cuda": "12.1" }, "node_hashes": { "._runner.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "runner.py": "309e32e16d4aea859a71ab8849bafef46382ce3b1afadcb896ba9c781eef5900", ".___init__.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "__init__.py": "cb71a8321d8238da3c0df658ad8f1a5577ab63d4d18b538be1fab98f8685c0b6", "._nodes.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "nodes.py": "286587fc721fa230b918adde3e085a8d931b12b13a5dda2a6f8edf47b15361f4", "._setup_worker.py": "23a10b56b67086f392f88c390f9edfe9e254e96a251d6bdc4ab82648791362dd", "setup_worker.py": "02c5e6c140dd6517f1bfcb3177b5a83320a7e8022efd8207c8c03e0d2e42d10f" } }, "telemetry": { "peak_gpu_mib": 8571.0, "mean_utilization_pct": 40.1, "mean_power_w": 173.066 }, "poster_seconds": 2, "notes": "The sampled crop retains the portrait\u2019s hair and face markings. Teeth are visible in the sampled speech frame; inspect the mouth edge and nearby cheek texture in the moving output.", "notes_scope": "Individual sampled-frame observation; not a temporal-artifact or lip-sync score.", "trial_kind": "Distinct input/settings run", "video_bytes": 270824, "preview_video": "/benchmarks/new-10-female-preview.mp4", "preview_sha256": "ba88ed3798fb0e8a2ee11a733f76fed6d1910cba70b409afa837b7c6bda00a90", "preview_bytes": 123971, "notes_review": "Reviewed the 2-second output frame against the original input." } ] }