Video inspect 400 video_inspect_transcribe_required: the fix

Sume video inspect returns 400 video_inspect_transcribe_required when a request carries language_code, segmentation, or duration_seconds and transcribe is not true. Those three fields only mean something to the speech-to-text step, so the API refuses them instead of ignoring them quietly.

This is spelled out in the video inspect docs, read 2026-10-05: send the STT fields only with transcribe: true.

The request that works

Add the flag, and give the hints you meant to give.

curl -X POST https://api.sume.com/v1/video-inspect \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: inspect-ko-001" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4",
    "transcribe": true,
    "language_code": "ko",
    "segmentation": {"mode": "sentence"},
    "duration_seconds": 90
  }'

What each hint does

• language_code is an STT hint such as en or ko. Without it, STT auto-detects.
• segmentation.mode: "sentence" adds gapless sentence segments[] shaped like caption lines. silence_split_seconds takes 0.2 to 3.
• duration_seconds is a billing hint. Without it Sume reserves 1 minute, and the hint maxes at 600 seconds.

Cost and a silent-clip trap

The transcript adds the public STT rate of $0.01 per audio minute on top of the Modal compute reservation for the inspect. A clip with no audio fails with inspect_source_has_no_audio, so check probe.has_audio with a frames: false call first when you are not sure.

The 400 is returned at admit, before anything is reserved, so hitting it costs nothing.