Video engines (face-swap, talking heads, character swap)

The video engines are grouped into families (see the catalog). Each takes a source plus a motion/target input:

Face swap - a source face (init_image) onto a target_video. Engines: FaceFusion, GHOST.
Talking head - a portrait (init_image) + audio -> a speaking clip. Engines: FantasyTalking, Hallo3.
Reenactment - a portrait + a driving_video performance. Engines: LivePortrait, FlashPortrait.
Character swap - a reference person + a source video: replace the WHOLE person (face, hair, body, clothes), keeping the motion. Engine: Wan-Animate.

These are currently preview (being wired to live GPU endpoints); the catalog status field flips to live as each is served. Call shape mirrors the generation API with the inputs above.

Did this answer your question? If not, contact our team and a real person will help.