Burn in captions from your own word timings with Sume (no STT)

If you already have a transcript with timings, send it to Sume's standalone captions endpoint as words (word level) or cues / segments (phrase level) and Sume burns that text at those times without running speech-to-text. Send only one of script_text, words, cues and segments. The job costs $0.20 for a clip of up to 60 seconds under the current fixed estimate, and the video must be a public HTTPS URL.

Details are from Video captions, read 2026-10-06.

Which field should I send?

The four caption inputs are mutually exclusive.

What happens on a silent clip?

Speech-to-captions on a clip with no speech fails with caption_no_speech and next_action: use_overlay_captions. That is not a policy rejection. Switch to cues and the job burns your text with no ASR.

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-cues-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/example/clean.mp4",
    "cues": [{"text": "Free shipping today", "start": 0.5, "end": 2.5}]
  }'

What does the code above need?

Replace the URL with a clip you control, and keep the shape of the cue: text, start and end in seconds. Sume rejects localhost, private-network, non-HTTPS and signed URLs, and it does not accept SRT uploads.