Talking Head Video

Turn a still avatar image + a narration track into a lip-synced talking-head MP4, fully offline. A thin wrapper drives JoyVASA (audio -> facial-motion diffusion) rendered by LivePortrait; you (the model) prepare a good avatar and narration, animate, review, and deliver.

This is the capstone of the local media set: it composes with the other skills so the whole thing - face, voice, music - is generated on-device.

flowchart LR
  Desc["avatar description (optional)"] -.->|image-gen| Avatar["avatar.png"]
  Text["script text (optional)"] -.->|voice-clone-narration| Audio["narration.mp3"]
  Avatar --> Anim["animate.py<br/>JoyVASA + LivePortrait (MPS)"]
  Audio --> Anim
  Anim --> Mp4["talking-head.mp4 (H.264 + AAC)"]
  Mp4 -.->|review-mp4| Check["frame/sync check"]