Avatar preview captions: stored at create, burned at generate-video
Preview stills are never captioned. Caption settings on the preview are stored, then applied at generate-video. What that means.
If you add captions to an avatar video preview, are the preview stills captioned? No. Preview stills are never caption-burned. The caption settings you send on create are stored with the preview and applied only when you call generate-video on it.
That is useful now that voices change often. ElevenLabs released Eleven v4 on 2026-09-28 (read 2026-10-04) and other vendors shipped new voices the same month. Keeping caption intent separate from the voice and the framing means you can approve a still once and still change what is spoken.
The flow
The preview family has four routes: POST /v1/avatar-video-previews, GET /v1/avatar-video-previews/:id, POST /v1/avatar-video-previews/:id/regenerate and POST /v1/avatar-video-previews/:id/generate-video. Create takes the same body as an avatar video: exactly one of script or video_inputs, plus optional product_image, scene, quality, aspect_ratio, title and captions.
Inline captions use the same four knobs as standalone captions: style, an optional font, a language hint and script_text. Styles are slam (default), punch, tiktok-green, korean-ad, plus the Hangul identities, per the preview guide.
What happens where
Read the table as a checklist before you pay for the full render.
| Step | Captions |
|---|---|
| Create preview | Stored for later; the still is never captioned |
| Regenerate still | Still has no captions; stored intent unchanged |
| generate-video | Captions burned into the clean final MP4 |
| Failure at the caption stage | The avatar job can still succeed with a clean primary video and captions.status=failed |
Limits to know
Estimated duration above 60 seconds is rejected for inline captions. A Korean script with slam, punch or tiktok-green is rejected with 400 caption_hangul_text_latin_style instead of being re-styled, because those faces render Hangul as tofu. Choose a Hangul style for Korean speech.
Because a caption failure soft-fails, check captions.status on the result before you ship, not only the job status. If it failed, you still hold a good clean video and can run a standalone video captions job on it.
Sources
Related posts
More in Sume Avatar 1.0
- Avatar video script too long? The 4 to 60 second rule
Sume Avatar 1.0 accepts scripts it estimates at 4 to 60 seconds. Estimate yours first, then shorten it or split it into separate videos before you submit.
- Griffin-style listening beats in a recorded avatar video: what works
Tavus Griffin nods and says mm-hm while it listens. A recorded Sume avatar video cannot listen, but it can hold a silent beat. Here is a preview-first test.
- Avatar validation failed: fix it, or skip verification
Searches for HeyGen avatar validation failed show people stuck on verification. On Sume an avatar create is a job you poll, and failures return a job error.
- Book launch avatar video: use the cover as the product image, cost
An author avatar introduces a book with the cover as the product image. A 30-second clip costs $5.82 at standard, $7.74 at plus and $17.40 at max.
Written by Sume