Three caption styles on one clip for $0.60 with source_caption_id

Burning three caption styles into one 45-second clip costs 3 x $0.20 = $0.60. The first job transcribes; the next two send source_caption_id instead of video_url, which reuses the stored source video and word timings, so speech-to-text does not run a second time. The docs say the price does not change on a restyle, because it is still a render.

The three jobs

Fixed estimate of $0.20 applies to videos of up to 60 seconds, per the video-captions docs read 2026-10-09.

Requests

The first call needs a public HTTPS video_url that Sume can fetch; localhost, private-network, signed and provider task URLs are rejected. The restyle call carries only the id and the new style.

Note that punch and tiktok-green do not support design overrides, while slam and the Hangul styles do.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: restyle-punch-001" \
  -d '{"source_caption_id":"vc_123","style":"punch"}'

Correcting words without a new transcription

If the transcript has a wrong word, send words with source_caption_id and only the text is corrected. You may send only one of script_text, words, cues and segments in a request.

Limits

• The $0.20 figure is for clips of at most 60 s under the current fixed estimate; check the catalog for the live price on longer files.
• Latin styles with Korean text return caption_hangul_text_latin_style.
• SRT upload is not supported; phrase-level text goes in cues.

Budget

Per-clip cost scales with the number of looks, not with transcription. Ten clips with three styles each is 30 jobs x $0.20 = $6.00. If you decide the style on a test clip first and then run the other nine in the chosen style, the cost is 3 x $0.20 + 9 x $0.20 = $2.40. Pick styles early; the restyle saves a transcription pass but not the render fee. Every media job follows the same lifecycle: submit with an Idempotency-Key, receive a job, poll GET /v1/jobs/:id/status until it is ready, then read GET /v1/jobs/:id/result. A retry with the same key does not queue a second job, so a network error during submit never doubles a charge.