echomimic

Why we included this project

EchoMimic turns a still portrait and a voice track into a video of the person speaking and moving their head. What sets it apart from plain lip-sync tools is the editable landmark-conditioning stage: on top of matching the audio, you can steer the expression, the pose, or individual facial movements. That flexibility means one codebase covers straightforward audio-to-video talking heads and the trickier audio-plus-landmark hybrids people hit in production. The repo includes inference scripts, accelerated models, pretrained weights for English and Mandarin, and Gradio and ComfyUI interfaces, so a small team can get it running instead of just reading about it. If you are building talking-avatar pipelines, virtual presenters, or dubbing work, this is a practical reference.