Korean caption fonts: 29 Hangul faces and each style's default

Sume's video captions API accepts 29 named Hangul fonts in the font field, from pretendard to east-sea-dokdo. If you send no font, the style keeps its own face: Pretendard for five styles and Do Hyeon for two. Any name outside the list is rejected, and the API never swaps in a replacement face.

Which face each style uses when you send no font

The docs give the default face per style. Only Hangul styles take a font at all.

The 29 names you can send

The list below is grouped by the feel the docs give each face. All of them use the SIL Open Font License 1.1 and the renderer includes them, so no font file has to be uploaded.

• Clean or workhorse: pretendard, noto-sans-kr, ibm-plex-sans-kr, gothic-a1, nanum-gothic, gowun-dodum
• Serif: noto-serif-kr, nanum-myeongjo
• Impact and thick: do-hyeon, black-han-sans, gasoek-one, bagel-fat-one, gmarket-sans
• Rounded and cute: jua, dongle, hi-melody, single-day
• Script and handwritten: nanum-pen, poor-story, gamja-flower, gaegu, cute-font, east-sea-dokdo, yeon-sung
• Display and retro: dunggeunmo, hahmlet, song-myung, stylish, sunflower

Rules that catch people

font is only for Hangul styles. If you send one of these faces with a Latin style such as slam, the API returns 400 with caption_font_requires_hangul_style. language never picks a font or a style; it only tells speech-to-text which language to expect.

The weight-shift and korean-ad styles animate the wght axis, and only Pretendard has that axis. On a static face the color and scale emphasis stay, but the weight travel is gone. If you want the motion, leave the face on Pretendard.

editorial-emphasis is the exception: its last-word emphasis line is Black Han Sans for every value of font, because the look is the contrast between two faces. Your font choice changes only the lead line.

Request example

This request keeps the korean-ad look and swaps the face to Gmarket Sans. A standalone caption job costs $0.20 for a clip of up to 60 seconds under the current fixed estimate, so a font change on a new job is a normal job price.

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

How to choose without guessing

Start from the style, not the font. The style decides phrasing, placement and motion; the font only changes the letterforms. Pick clip-wipe or black-outline for dense talking-head speech, and change the face only if the brand needs it.

A practical test is to burn one 15-second clip three times and compare faces on a phone. Use source_caption_id to restyle the same job so you do not transcribe again. Each restyle is a normal render at the standalone job price. Treat the feel column in the docs as a starting point: 'Impact' faces such as Black Han Sans and Gasoek One read well at small sizes, while script faces such as Nanum Pen need larger text and a quiet background.

• Pretendard is the only face with the wght axis, so it is the one that gives weight-shift and korean-ad their weight travel.
• Fonts are bundled in the renderer, so there is nothing to upload and nothing to license separately under the OFL 1.1 terms the docs state.
• Unknown names fail at request time; no fallback face is used.