Restyle burned-in captions without transcribing twice

To change the look of captions you already made, send source_caption_id and a new style instead of a video_url. Sume reuses the earlier caption's source video and the word timings it already holds, so no second speech-to-text runs. The docs are clear that billing is unchanged: a restyle is still a render.

That makes it the right tool for A/B testing looks, and the wrong tool for saving money.

What does the request look like?

The body is a source_caption_id and the new style, for example { "source_caption_id": "...", "style": "black-outline" }. You get the id from the earlier caption job. Pass words alongside it only when you need to correct the wording.

When should you use it?

Mind the script rule. Latin styles reject Korean text with a 400, so a restyle from a Hangul style to slam will fail for a Korean clip. Choose another Hangul style.

• Comparing two or three styles on the same clip before choosing a house look.
• Moving a Korean clip from one Hangul style to another, where the faces differ.
• Fixing a single misheard word without re-timing everything, by sending corrected words.
• Producing platform variants that differ only in caption design.

What about translated text?

A restyle keeps the original wording unless you correct it with words. For a translated version you would send new cues on the video instead, as in the SRT post, and check terms first with the glossary check.

A standalone caption job is priced at $0.20 for videos up to 60 seconds under the current estimate, and a restyle costs the same as a render does, so plan the number of variants. The full field list is in Video captions.

How do you keep track of variants?

Store the caption id, the style and any design overrides for every variant you render. When a client picks one, you can reproduce it, and when a style changes you know which jobs to redo. Remember that each restyle is billed as a render, so decide on two or three candidates first and not ten.

If you compare more than two styles, put the same cue on each at the same frame in a contact sheet. Differences in weight and placement show up faster side by side.