Synthesia animated scene preview vs Sume first-frame stills

Synthesia's update lets you preview one scene with a fully animated avatar from the editor. Sume's preview does not animate: POST /v1/avatar-video-previews creates a first-frame still, so you can approve composition before starting the full talking-video render. Motion and lip movement are judged only on the final render.

Synthesia's wording is from its updates page; Sume's from Avatar video previews, read 2026-10-01.

What does the Synthesia preview do?

The updates page describes previewing an avatar's animations for a single scene from the editor, so you can review and iterate on avatar performance without generating or re-generating the whole video.

What does a Sume preview show?

The docs say previews generate the first-frame still stage without starting the full talking-video render. For multi-scene video_inputs you get one still per scene. Captions are stored on create and applied only at generate-video; preview stills are never caption-burned.

How do I go from preview to final video?

Four routes cover the flow: create POST /v1/avatar-video-previews, read GET /v1/avatar-video-previews/:id, redo stills with POST .../regenerate, then POST .../generate-video. The create body matches Avatar Video fields, with exactly one of script or video_inputs; quality defaults to plus. See first-frame previews for avatar video.

When is a still enough?

Use it to check framing, product placement and scene setup. If what you need to review is gesture or speech performance, a still cannot show it; that needs the rendered clip. For a direct render without the preview stage, use Generate avatar video.