Captions app pricing is iOS-only: caption videos from code instead

If you need captions from a script or a server, the Captions app's pricing page does not help, because it says its prices and features reflect iOS plans only and it does not mention an API. Sume's caption endpoint takes a public video URL and burns the captions in for $0.20 per accepted job, for videos of up to 60 seconds.

What the Captions page lists

The free plan offers basic editing such as trim, transitions and captions but no generative features. The page does not state export resolution, watermark status or API access.

What the Sume caption endpoint does

You can also send your own words or cues with start and end times, and Sume then skips speech-to-text, which is the route for silent clips.

• Input: a public HTTPS video_url, plus optional style, font, language and script_text.
• Default style is slam for Latin text (Korean text resolves to black-outline), with other looks such as punch, weight-shift, highlight, pill-karaoke, clip-wipe and editorial-emphasis.
• Price: $0.20 reserved and captured per accepted standalone job, for videos up to 60 seconds.

The arithmetic, with the caveat

At $0.20 a clip, $24.99 pays for 124 caption jobs. That is not a comparison with Max, whose credits cover generative edits as well as captions, and the page does not say how many credits a caption costs. The honest framing is access: an iOS subscription is for a person editing on a phone, and a per-job API is for a pipeline. Pick by who presses the button.

Restyling without paying for speech-to-text twice

To change the look of an already captioned video, send source_caption_id and a new style instead of the video URL. Sume reuses the source video and the stored word timings, so transcription does not run again. The docs say the price does not change, because a restyle is still a render. Budget each restyle as another $0.20 job, and decide the style before the first render when you can. If you only need a different language hint, note that language tells speech-to-text what to expect and never selects a style or font. Korean text sent to a Latin style is refused with a 400 rather than rendered as empty boxes, so the failure is loud and costs you nothing to fix.