Video inspect transcript: the one-minute reserve and duration_seconds

Asking for a transcript through POST /v1/video-inspect with transcribe: true runs Sume STT 1.0 at $0.01 per audio minute. If you omit duration_seconds, Sume reserves one minute; if you pass it, it is a hint with a maximum of 600 seconds. Probe and stills are not billed, so the transcript is the only part that touches your balance.

All figures are from the video inspect docs, read 2026-10-02; the docs say to confirm the live rate in GET /v1/catalog.

What does duration_seconds do?

It is an optional hint valid only with transcribe: true; sent without it, the call returns 400 video_inspect_transcribe_required. Its job is to size the reserve. The docs state the omitted-hint reserve (one minute) and the 600-second maximum for the hint, and do not describe what happens to a clip longer than the hint, so keep the hint close to the real audio length.

How do I keep the cost clear?

Probe first with frames: false to read the clip length and probe.has_audio. A silent clip fails as inspect_source_has_no_audio. At the stated rate, a 5-minute talk is 5 minutes of audio at $0.01. Set language_code (for example en or ko) when you know it, or omit it for auto-detect.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: transcript-300s-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "frames": false,
    "transcribe": true,
    "duration_seconds": 300,
    "language_code": "en"
  }'

Which parts of an inspect are free?

Read the finished inspect with GET /v1/video-inspect/:id, or poll the job with jobs_wait over MCP.