HeyGen photo avatar hand gestures vs Sume Motion Control

For HeyGen Avatar V photo avatars, a video can now render straight from the photo, but motion_prompt still requires an animation reference, so hand gestures need an eligible reference look. On Sume, a prompt steers the scene of an Avatar Video, while prompted-style body motion from a reference is a separate Motion Control route.

What changed in HeyGen's August 2026 note?

The changelog says that when an Avatar V photo avatar's group has no eligible digital-twin or curated reference look, video creation renders directly from the photo instead of rejecting the request. That works when reference_look_id is omitted and no eligible look can be selected automatically. It then says motion_prompt still requires an animation reference, and to supply an eligible reference look when you need prompted body motion or hand gestures. (Read 2026-09-30.)

How does Sume steer an avatar video?

Avatar Video takes scene: { "type": "prompt", "prompt": "..." } for scene direction, and a photo scene as another option. That directs the setting, not limb motion. Current execution supports one resolved avatar per final video; see Avatar videos.

Where does body motion come from on Sume?

The OpenAPI file has a face-control alias, createAvatarV1MotionControl, described as a face-control alias of POST /v1/kling/3.0/motion-control with the same body; created jobs store the public model id kling/3.0/motion-control. On the hosted MCP the tool is kling-motion-control_create, listed in the MCP docs. Motion comes from a reference you supply. Read AI avatar hand gestures and dance moves for the setup.

Which route should I use?

If you need a person speaking a script in a chosen scene, use Avatar Video. If you need specific gestures or dance, use Motion Control with a reference clip. They are separate jobs with their own limits, so check each one before combining them.