Captioning a 3-minute Short: Sume prices 60 s, so plan 3 jobs

Budget $0.60 in caption fees for a 180-second YouTube Short, and verify it in GET /v1/catalog first. The Sume captions page says each accepted standalone caption job reserves and captures $0.20 for videos of maximum 60 seconds under the current fixed estimate. It does not say what a longer video costs, so the safe plan is three 60-second parts. YouTube's help page says Shorts can run up to 3 minutes, so the full Short fits the platform.

The arithmetic

Rates are from the Sume docs; the three-minute limit is from the YouTube page.

Steps

The captions API takes a public HTTPS video URL, unlike trim and Timeline, which take media.sume.com URLs. Check that the URL of your artifact is reachable that way before you build the batch, because the docs do not say it is.

• Trim three parts with start 0, 60 and 120 and duration 60.
• Call POST /v1/video-captions for each part with a style such as black-outline and a language hint.
• If a restyle is needed, send source_caption_id instead of video_url. The docs say the word timings are reused and the price does not change.
• Read each result and review the text.
• Publish the three parts as a series, or assemble them in your own editor.

A silent part fails with caption_no_speech, so check the audio first with video_inspect (frames: false, then read probe.has_audio).

Is the split worth it?

Splitting costs a little more in money and a lot more in handling, so decide whether the three-part series is a good creative choice before you decide it is a good cost choice. A 180-second gift guide told in three 60-second Shorts can be easier to scan and gives you three posts, each with its own hook. A single 180-second Short keeps the viewer in one place.

If you want a single file, you can pay once for the full-length caption job and read the price Sume returns for it. The docs call the $0.20 a fixed estimate for up to 60 seconds, so a longer clip may be priced differently or refused, and a dry_run-style check is not documented for this endpoint. Test with one real clip and read the charge before you queue a hundred.

Whichever you pick, keep the cue text short. Three lines of four words each read better than one long sentence.

What Sume does not do

Sume does not document a single caption price for a 180-second video, and I did not test one. The tool also burns captions into the frames, so it does not give you YouTube's separate caption track.