avatar-image-to-video_create: choosing H3 Max lip sync over Fabric

In Sume MCP, avatar-image-to-video_create is the only talk tool, and its model field chooses the lip sync model: veed/fabric-1.0 is the default and minimax/h3-max/lip-sync is the explicit alternative. generate_video refuses the lip sync id with wrong_tool. The Agents @ catalog lists it as @minimax-h3-max-lip-sync next to MiniMax H3 Max.

From docs/api/minimax-h3-max-lip-sync.md on origin/main b28f3b0, read 2026-10-13.

What does the strict body refuse?

On the HTTP routes the URL is the model, so the body must not carry model, endpoint or provider_endpoint. Exactly one visual source is required: image_url (public HTTPS, aspect ratio 0.4 to 2.5) or a ready avatar by id or handle. Audio is a Sume-hosted audio_url of 10 MB or less, with a required duration_seconds from 5 to 14.8.

Which model should the agent pick?

Default to Fabric. Use H3 Max only inside the 5 to 14.8 second window, and use one lip sync model per run, since the model sets the assemble output.fps lock (Fabric is 25 fps; H3 Max must be measured on the first real clip).