SadTalker Google Colab Guide: Free T4 GPU Setup, Error Fixes & Notebook

!nvidia-smi

!apt-get update -qq && apt-get install -y -qq ffmpeg git wget
!git clone https://github.com/OpenTalker/SadTalker.git /content/SadTalker
%cd /content/SadTalker

%%bash
# Auto-accept Conda Terms of Service for non-interactive execution
export CONDA_PLUGINS_AUTO_ACCEPT_TOS=true

# Install Miniconda quietly if not already installed
if [ ! -d "/content/miniconda3" ]; then
  wget -q https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
  bash Miniconda3-latest-Linux-x86_64.sh -b -p /content/miniconda3
  rm -f Miniconda3-latest-Linux-x86_64.sh
fi

# Create dedicated Python 3.8 environment
source /content/miniconda3/bin/activate
if ! conda info --envs | grep -q "sadtalker"; then
  conda create -n sadtalker python=3.8 -y
fi

# Activate and install validated compatible packages
source /content/miniconda3/bin/activate sadtalker
pip install --quiet --upgrade pip
pip install --quiet numpy==1.23.5 torch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 \
  facexlib==0.3.0 gfpgan insightface onnxruntime moviepy \
  opencv-python-headless imageio[ffmpeg] yacs kornia gtts \
  safetensors pydub librosa

# Apply BasicSR Torchvision compatibility patch
sed -i 's/from torchvision.transforms.functional_tensor import rgb_to_grayscale/from torchvision.transforms.functional import rgb_to_grayscale/' /content/miniconda3/envs/sadtalker/lib/python3.8/site-packages/basicsr/data/degradations.py 2>/dev/null || true

echo "Conda environment configured and ready!"

%%bash
cd /content/SadTalker
mkdir -p ./checkpoints ./gfpgan/weights

# Core SadTalker models
wget -nc https://github.com/OpenTalker/SadTalker/releases/download/v0.0.2-rc/mapping_00109-model.pth.tar -O ./checkpoints/mapping_00109-model.pth.tar
wget -nc https://github.com/OpenTalker/SadTalker/releases/download/v0.0.2-rc/mapping_00229-model.pth.tar -O ./checkpoints/mapping_00229-model.pth.tar
wget -nc https://github.com/OpenTalker/SadTalker/releases/download/v0.0.2-rc/SadTalker_V0.0.2_256.safetensors -O ./checkpoints/SadTalker_V0.0.2_256.safetensors
wget -nc https://github.com/OpenTalker/SadTalker/releases/download/v0.0.2-rc/SadTalker_V0.0.2_512.safetensors -O ./checkpoints/SadTalker_V0.0.2_512.safetensors

# GFPGAN & face enhancement models
wget -nc https://github.com/xinntao/facexlib/releases/download/v0.1.0/alignment_WFLW_4HG.pth -O ./gfpgan/weights/alignment_WFLW_4HG.pth
wget -nc https://github.com/xinntao/facexlib/releases/download/v0.1.0/detection_Resnet50_Final.pth -O ./gfpgan/weights/detection_Resnet50_Final.pth
wget -nc https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.4.pth -O ./gfpgan/weights/GFPGANv1.4.pth
wget -nc https://github.com/xinntao/facexlib/releases/download/v0.2.2/parsing_parsenet.pth -O ./gfpgan/weights/parsing_parsenet.pth

from google.colab import files
import os

print("=== Select Source Portrait Image (PNG, JPG, WEBP) ===")
uploaded_img = files.upload()
if uploaded_img:
    img_name = list(uploaded_img.keys())[0]
    source_img_path = f"/content/{img_name}"
    print(f"Source Image loaded: {source_img_path}")
else:
    source_img_path = "/content/SadTalker/examples/source_image/full_body_1.png"
    print(f"No image selected. Using demo portrait: {source_img_path}")

print("\n=== Select Driven Speech Audio (WAV, MP3) ===")
uploaded_aud = files.upload()
if uploaded_aud:
    aud_name = list(uploaded_aud.keys())[0]
    driven_audio_path = f"/content/{aud_name}"
    print(f"Driven Audio loaded: {driven_audio_path}")
else:
    driven_audio_path = "/content/SadTalker/examples/driven_audio/bus_chinese.wav"
    print(f"No audio selected. Using demo audio: {driven_audio_path}")

# Persist environment variables for bash execution
with open("/content/sadtalker_inputs.env", "w") as f:
    f.write(f"export SOURCE_IMAGE='{source_img_path}'\n")
    f.write(f"export DRIVEN_AUDIO='{driven_audio_path}'\n")

print("\nInputs saved successfully! Proceed to generation.")

%%bash
source /content/miniconda3/bin/activate sadtalker
cd /content/SadTalker
source /content/sadtalker_inputs.env

echo "Processing Source Image: $SOURCE_IMAGE"
echo "Processing Driven Audio: $DRIVEN_AUDIO"

python inference.py \
  --driven_audio "$DRIVEN_AUDIO" \
  --source_image "$SOURCE_IMAGE" \
  --result_dir ./results \
  --size 512 \
  --still \
  --preprocess crop \
  --enhancer gfpgan

import glob
import os
from IPython.display import HTML, display
from base64 import b64encode
from google.colab import files

mp4_files = glob.glob("/content/SadTalker/results/*.mp4") + glob.glob("/content/SadTalker/results/*/*.mp4")
if mp4_files:
    latest_video = max(mp4_files, key=os.path.getctime)
    print(f"Generated Video: {latest_video}")

    # Render HTML5 video player
    video_data = open(latest_video, 'rb').read()
    data_url = "data:video/mp4;base64," + b64encode(video_data).decode()
    display(HTML(f"""
    <div style="text-align: center; margin: 16px 0;">
      <video width="512" height="512" controls autoplay loop style="border-radius: 12px; box-shadow: 0 8px 24px rgba(0,0,0,0.15); max-width: 100%;">
        <source src="{data_url}" type="video/mp4">
        Your browser does not support the video tag.
      </video>
    </div>
    """))

    # Download to local device
    print("Downloading video to your local device...")
    files.download(latest_video)
else:
    print("No generated video found in /content/SadTalker/results/")