Localize One Ad to Korean: Burned-In Captions With Hangul Styles

The short answer

Send the video to POST /v1/video-captions with language: "ko" and a Hangul style such as korean-ad or black-outline. If you send Korean text to the Latin styles slam, punch or tiktok-green, the API returns 400 caption_hangul_text_latin_style, because those fonts have no Hangul glyphs and would draw empty boxes. A standalone caption job is $0.20 for a video up to 60 seconds, per the docs page read on 2026-10-09.

Pick the style

Docs list six Hangul identities for Korean speech and an ad look. If you do not send style, Korean text resolves to black-outline. korean-ad has to be chosen on purpose; it is CapCut-style karaoke, one short phrase at a time in the lower third, with the spoken word at a heavy weight. The language field is only a speech-to-text hint and does not choose the style or the font.

Fonts, scripts and restyles

The font field only applies to Hangul styles, and the faces are listed in the docs: Pretendard for most identities, Do Hyeon for black-outline and clip-wipe, and a longer list including Black Han Sans, Jua and Gmarket Sans. Sending a Hangul face to a Latin style returns 400 caption_font_requires_hangul_style. The weight-shift and korean-ad motion animates the weight axis, which only Pretendard has.

If you have the translated script, send it as script_text. Sume keeps the speech-to-text word timings as the clock and aligns your text to them; a mismatch gives script_alignment_mismatch, and the suggested next step is to simplify the script or omit it. To restyle a finished render, send source_caption_id and a new style: Sume reuses the stored word timings, so speech-to-text does not run again, and the price is still one render.

A workable routine for a campaign is to caption once in black-outline, show the client, then restyle with source_caption_id into korean-ad or clip-wipe if they want more motion. The restyle is still a render and costs the same $0.20 for a video up to 60 seconds, but it skips the second speech-to-text run. Check the live price in GET /v1/catalog before a long campaign.

• Docs describe the caption job as transcribing the speech in the video; they do not describe translation. If the audio is English, supply Korean text as cues or words, or make a Korean voice track first.
• Use cues (text, start, end) for a silent clip; speech-to-captions on a silent clip fails with caption_no_speech.
• Have a Korean speaker review the final text; auto-timing is not a proofread.