YouTube AI label and captions: burned-in subtitles on a real clip

What YouTube exempts

YouTube's page on altered or synthetic content lists caption creation among the uses that do not need the AI disclosure, next to beauty filters, color adjustments and personal voice cloning (YouTube Help, read 2026-10-05). The ask is about realistic content that makes a person appear to say or do something they did not.

What that means for subtitles

If the footage is real and the only change is burned-in text, the page's own list says the caption step does not trigger the Yes. If the footage also includes a generated face or voice, that part still falls under the disclosure rule, whatever the captions are.

The Sume side

Standalone captioning is POST /v1/video-captions and costs $0.20 for a video up to 60 seconds. It accepts a design object. Captions made inline on an avatar video accept style, font, language and script_text, but not design. Korean text on a Latin style such as slam returns 400 caption_hangul_text_latin_style.

Burn captions into a clip

curl -X POST https://api.sume.com/v1/video-captions \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: captions-real-clip-001" \
  -d '{"video_url":"https://media.sume.com/artifacts/artf_demo/talk.mp4","style":"punch","script_text":"Three things changed in this release."}'

Before you publish

Check the video-captions docs for the exact body. Then decide the AI use answer from what is on screen, not from the captions.