Bilingual captions crowd a vertical Reel: lay out two lines

Show two languages at once by cutting each line to 4 words or fewer, setting them in the clear middle of the frame, and putting the second language in a smaller or lighter weight. Instagram's Edits app now places captions in two languages on one video, but two full tracks are more text than a vertical frame holds.

Storyboard18 reports that Edits translates captions into a second language across 15 languages, from English and Spanish to Hindi, Korean, Japanese and Kannada (Storyboard18, read 2026-10-07). GrowthViral warns that the feature adds text only, that machine translation stumbles on names and slang, and that two full tracks can overcrowd a vertical video (GrowthViral, read 2026-10-07).

A text budget that fits

A 1080 by 1920 frame has a busy bottom and top, because the Instagram interface sits over both. A two-line stack works best in the middle band: first language larger, second smaller. Use these as starting limits and look at a test frame.

Doing the same layout in a Sume caption job

Sume's video captions accept a design block that overrides one style per request. The limits in the API for the fields below are: phrasing.max_words 1 to 12, phrasing.max_chars 4 to 60, phrasing.pause_seconds 0.05 to 3, typography.font_size_ratio 0.01 to 0.4, typography.safe_width_ratio 0.3 to 1 and placement.anchor_ratio 0.05 to 0.95 (the centre of the line as a fraction of the frame height). A value outside its range returns a 400 before any render, so a wrong look costs nothing.

A caption job burns one language track from speech or from cues. For two languages in one frame, send cues whose text holds both lines, written for short cards, and give the second language its own line in each cue. Test one clip first, because how a second line renders depends on the style you choose.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: bilingual-test-01" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/clean.mp4",
    "style": "black-outline",
    "cues": [{ "text": "Hola a todos\nHello everyone", "start": 0.2, "end": 2.0 }],
    "cues": [{ "text": "Hola a todos\nHello everyone", "start": 0.2, "end": 2.0 }],
    "design": {
      "phrasing": { "max_words": 4, "max_chars": 24 },
      "placement": { "anchor_ratio": 0.5 }
    }
  }'

The Korean rule

If one of the two languages is Korean, pick a Hangul style. slam, punch and tiktok-green use Latin faces with no Hangul glyphs, and the API returns 400 (caption_hangul_text_latin_style) instead of burning boxes. black-outline is the safe default. Each accepted caption job is $0.20 for a clip of up to 60 seconds.

Check the layout on a single frame

Do not judge a bilingual layout from a settings screen. Burn a test caption, then pull a still with video frames at a moment when both lines are on screen, and look at it at phone size. If the two lines touch the bottom interface, raise anchor_ratio; if they touch each other, lower max_chars. Two changes and one test frame are cheaper than a full re-render of a long Reel.

When one language is enough

A second caption track is not always an improvement. If your audience understands the spoken language, a clean single track reads faster. Use two when the audience is genuinely mixed, such as a travel Reel with locals and visitors, and keep the translated line shorter than the original so the eye has an obvious place to rest.