Transcribe a video in one call: video-inspect with sentence segments

If you only need the text of a hosted video, call POST /v1/video-inspect with transcribe: true and skip the detach step. The transcript costs $0.01 per audio minute, billed at the STT 1.0 rate on top of the inspect's own compute, which Sume captures at actual container seconds. The maximum duration hint is 600 seconds. See Video inspect.

Request

Set frames: false to skip stills if you only want words. segmentation.mode: "sentence" also returns sentence segments[] with no gaps, in the shape of caption lines.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: transcribe-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "frames": false,
    "transcribe": true,
    "duration_seconds": 240,
    "language_code": "en",
    "segmentation": { "mode": "sentence", "silence_split_seconds": 0.8 }
  }'

Hints and guards

Detach or inspect?

Inspect is one call when the source is a hosted video under ten minutes. Detach first when you need the audio as a file too, when the track is longer than the 600-second hint, or when you want to split one track into many ranges. A four-minute clip at $0.01 per minute adds $0.04 of transcript.

Read transcript.text, words[] and, with segmentation, segments[] from the result; the transcript also carries the audio_url it used.

Sync by default

Video inspect defaults to mode: sync and waits at most 30 seconds. A short clip comes back as 200 with the transcript; otherwise you get 202 and a job id, and you read the same object later with GET /v1/video-inspect/{id}. A source can be up to 1,800 seconds, while the duration_seconds hint that sizes the transcript reserve tops out at 600.