Korean captions on a small phone: clip-wipe or black-outline?

For Korean speech on a small phone screen, start with clip-wipe. The Sume docs call it "the most clear style at small phone sizes". If you do not want to choose, send no style for Korean text: the docs say the text resolves to black-outline, which they call the safe default. Both use the Do Hyeon face unless you set font. A standalone caption job costs $0.20 whichever style you pick.

The six Hangul identities

These are the styles that the docs list for Korean speech (talking head, UGC, creator voiceover). They group words into phrase cards and burn the transcript as written, with no case folding.

Why the face matters on a phone

The docs add a separate style, korean-ad, for the CapCut-style ad karaoke look. It resolves from style: "korean-ad" only. If you send no style, you do not get it. weight-shift and korean-ad animate the wght axis, and only Pretendard has that axis. If you set font to a static face such as do-hyeon or black-han-sans, those two styles keep their color and scale emphasis but lose the weight travel.

So the practical split is this. Use clip-wipe or black-outline when readability at a small size is the goal. Use weight-shift or korean-ad when you want the weight animation and keep the Pretendard face.

Three rules that cause a 400

• language is only a speech-to-text hint. It never selects the style or the font. Sending language: "ko" without a Hangul style does not make the render Korean-safe.
• Korean text on slam, punch or tiktok-green returns 400 with caption_hangul_text_latin_style. Those faces have no Hangul glyphs.
• A Hangul font with a Latin style returns 400 with caption_font_requires_hangul_style.

Choosing between the two safe styles

black-outline and clip-wipe both use Do Hyeon, so the face is the same and the difference is the motion. black-outline is the quiet choice: the phrase sits in the middle of the frame and the fill sits on a thick outline. clip-wipe adds a left-to-right wipe on each word, which gives the eye a point to follow when the text is small. If your viewers watch with the sound off, the wipe helps. If your clip already has a lot of motion, the static outline competes less with it.

Test both on one clip. A restyle with source_caption_id reuses the word timings, so the second look does not run speech-to-text again, and each job is $0.20.

A request to start from

Send the style and the language hint. Add font only if you want a face other than the style default. If you want a fixed brand color, use design.colors.active; the docs show that example on black-outline.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ko-clip-wipe-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/clean.mp4",
    "style": "clip-wipe",
    "language": "ko"
  }'