Recreate a winning ad's structure for your product with Sume
sume-recreate keeps a reference video's scene order, beat timing and caption rhythm, casts a new presenter and generates every shot fresh. How to call it.

sume-recreate rebuilds the structure of a reference video for your product: it keeps the scene order, beat timing and caption rhythm, casts a new presenter, and generates every still, clip, audio track and caption from scratch. Call it at the sume handle with your packshot attached and the reference URL, or clip, named in the instruction.
What it keeps and what it makes new
| Kept from the reference | Made from scratch |
|---|---|
| Scene order | Every still, one per beat |
| Beat timing | Every clip, animated from the accepted still |
| Caption rhythm | The audio track and the captions |
| The music-or-voice choice | A new presenter |
Call it
Send POST /v1/formats/sume/sume-recreate/runs with formats:write, an Idempotency-Key and your product packshot in attachments. In the instruction, give the reference link and describe your product. The catalog brief assembles the beats on the timeline and adds the caption layer last, so you do not script those steps.
The run can take a while because it builds a still per beat, animates each, assembles them and adds captions. Use a webhook on the run instead of holding a connection open, and poll status_url as a fallback. A canceled run sends no webhook, so a poll covers that case.
Use references you have the right to rework
The Format makes new media from your product, but the structure comes from someone's video. Use your own past winners or references you are licensed to rework, and keep the result clearly yours.
Make it a batch
One reference with ten products is ten items on one queue. Put the product row in input, give each item a cap, and set a concurrency your wallet can hold. For one product against ten references, put the reference URL in each item's instruction. Read each child receipt before you rank outputs.
Limits worth knowing
- Check
output_errorand the receipt before you count a run as delivered. - The Format's instruction is composed after the Format body, and your
instructionwins where they disagree. - The default spend cap is the Format's, and $400 if none was set. Lower it per run.
previous_run_idcontinues a run, which suits fixing one beat without redoing the rest.
Sources
Related posts
More in Formats
- Render a Short and wait for it in one request? Timeline sync mode
Timeline 1.0 can answer a Short render in one request when mode is sync and the job finishes within 30 seconds. Otherwise it returns 202 and you poll.
- Season outline as nested input: one key, 64-key limit, 2 MiB
Put a whole season outline under one key of a Format run's input. Only top-level keys count toward 64, and the body can be up to 2 MiB, so 40 episodes fit.
- Shorts series sequential playback: fade out or hard cut each episode?
YouTube plays Shorts series episodes in order. Pick a hard cut or a short fade at the end of each one, with the Timeline 1.0 fade limits.
- Bulk queue item error: format_run_canceled vs format_run_failed
In a Sume bulk queue, a failed child reads format_run_failed and a canceled child reads format_run_canceled. The real reason is in the child run receipt.
Written by Sume