Instagram's Devanagari caption fonts: what Sume burns for Hindi

Instagram added font script options for Devanagari and Bengali-Assamese, so creators can style captions in Hindi, Marathi, Bengali and Assamese inside the app. Sume's video captions docs describe Latin faces and 29 Hangul faces; they document no Devanagari or Bengali face, so burning Hindi or Bengali captions with Sume is untested territory you should check on one clip first.

The Instagram facts come from Meta's announcement, read on 2026-10-03. The page says Meta AI translation was rolling out to everyone on January 16, 2026.

What did Instagram announce?

The page says Meta AI now translates, dubs and lip-syncs reels in Bengali, Tamil, Telugu, Kannada and Marathi, joining Hindi, English, Spanish and Portuguese. It says the tool preserves the sound and tone of the creator's voice and that lip sync is optional.

For text, creators gain Devanagari and Bengali-Assamese script options for captions and text styling. To reach them, open Text in the editing timeline, tap the Aa icon, and filter fonts by language by swiping down on the all-fonts tab. The page says the Android rollout began in the announcement period.

What font faces does Sume document?

The video captions page is specific. Latin styles (slam, punch, tiktok-green) draw in Latin display faces. Hangul styles ship with a list of open-licensed Korean faces, including Pretendard and Do Hyeon. It states that every face listed is under the SIL Open Font License 1.1 and ships with the renderer, and that any other font name is rejected rather than substituted.

Nothing on that page lists a Devanagari or Bengali face. That does not prove the job will fail, and it does not prove it will work. It means Sume has made no promise, so a tofu-box result is a real possibility, the same failure the page describes for Korean on a Latin style.

How do you test Hindi captions without wasting money?

Run one short job before a batch. Pass the Hindi line as a single cue so no speech-to-text can change the wording, then look at the output frame by eye. The page says cues take text, start and end in seconds and skip speech-to-text.

A standalone job reserves $0.20 for videos up to 60 seconds under the current fixed estimate, so a test costs about that, and the docs say to confirm live pricing in GET /v1/catalog.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: hindi-glyph-test-001" \
  -d '{"video_url": "https://media.sume.com/artifacts/artf_demo/clean.mp4", "style": "slam",
       "cues": [{"text": "\u0928\u092e\u0938\u094d\u0924\u0947", "start": 0.2, "end": 1.6}]}'

What should you do if the test shows empty boxes?

Treat that as the answer, not a bug to retry. Burning the caption inside Instagram's own editor, where Meta's font filter works, is the safe route for those scripts. Keep Sume for the parts it documents: trimming, joining audio and captions in Latin or Hangul.

• Do not retry the same job; a font gap does not fix itself on a second run.
• Use the in-app Aa font filter for Devanagari and Bengali-Assamese text.
• Keep a clean, uncaptioned master so the app can caption it.
• Report the result in your own notes; Sume's docs are the source of truth for support.