Korean ad captions on a 30-second spot: korean-ad style, $0.20

The settings

For Korean speech, send style korean-ad and language ko to POST /v1/video-captions. The docs describe korean-ad as CapCut-style Hangul karaoke: one short phrase at a time in the lower third, with the spoken word changing to a heavy weight. Korean script_text is supported. A 30-second spot costs $0.20 as a standalone caption job.

What each field does

If you do not send style, the text decides: slam for Latin and black-outline for Korean. language only tells speech-to-text which language to expect; it does not select the typeface. If you leave it out, speech-to-text finds the language by itself.

Avoiding tofu boxes

The docs warn about missing glyphs: a video full of tofu boxes costs the same as a good one. Use a Hangul-capable style or font identity rather than a Latin-only face for Korean text, and review the first frames of the output.

If the clip is a silent animation with authored text, send cues with text, start and end in seconds instead; speech-to-text does not run for that path.

Batch cost

Twenty 30-second Korean spots at $0.20 each is $4.00. Adding a second style variant per spot doubles the cost to $8.00 for the render, though a restyle with source_caption_id skips the second transcription. Reference: Video captions.

Reviewing the output

Check three frames: the first caption, a mid-sentence caption and the last. You are looking for missing glyphs, lines that overflow the lower third and emphasis landing on the wrong word. If a word is wrong, fix it with words on a restyle instead of re-running everything.

Write script_text exactly as it is spoken, so the alignment has nothing to guess.

Mixed-language spots

If your spot mixes Korean with English product names, keep language as ko and supply script_text so the English names are burned exactly as written. The docs say the default style depends on the text, so a script that is mostly Latin will default to slam unless you set korean-ad yourself.