SadTalker
OpenTalker/SadTalker · Python · Apache-2.0
Talking-head video from one photo plus audio
SadTalker remains a practical way to animate a still portrait from a voice track. Newer avatar stacks exist, but this repo is still widely forked and documented.
Score
Setup at a glance
git clone https://github.com/OpenTalker/SadTalker.git && pip install -r requirements.txt Deep dive
SadTalker turns a single portrait photo plus an audio clip into a talking-head video — still the staple tool when an avatar needs to speak, from virtual streamers to product demos.
It generates stylized 3D motion coefficients (head pose, expression) from the audio and renders photo-realistic frames, which is why results look more natural than naive warping approaches.
The license moved to Apache 2.0, it integrates with stable-diffusion-webui as an extension, and a local Gradio demo (app_sadtalker.py) plus the official project site round out the tooling.