STT 1.0 or video inspect transcribe: two ways to transcribe on Sume

Sume has two transcription paths that look alike. POST /v1/stt-1.0/transcribe is the standalone speech-to-text model. POST /v1/video-inspect with transcribe: true runs the same Sume STT 1.0 on the audio of one video clip. Both price the transcript at $0.01 per audio minute (read 2026-10-07 from the video inspect docs and the repo's catalog), and both can return word timings and sentence segments. The differences are in what you hand them and what else gets billed.

Side by side, read 2026-10-07

Choose by what you already have

If you hold a recording, call STT: one job, one rate, nothing else on the invoice. If you hold a video and also want its probe and stills, call inspect with transcribe: true, and accept that the inspect's compute is billed beside the minute rate. If you hold a long video and only want the words, detach the audio first (POST /v1/audio-detach, $0.01) with format: wav, channels: mono and sample_rate: 16000, then send that file to STT.

Either way, ask for sentence segments with segmentation: {"mode": "sentence"} when you want caption-shaped lines. The transcript output is text, language_code, words[] as {word, start, end} and segments[] as {index, text, start, end, duration_seconds}.

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-talk-001" \
  -d '{"video_url": "https://media.sume.com/artifacts/artf_demo/talk.mp4", "frames": false, "transcribe": true, "duration_seconds": 120, "segmentation": {"mode": "sentence"}}'

What a 10-minute job costs

A 10-minute recording through STT 1.0 is $0.10. The same minutes through inspect are $0.10 of transcript plus the compute of the inspect job. If both are billed to the cent, STT is the cheaper path for pure transcription by that compute amount. If the clip is already a video in your library, inspect saves you the detach step and gives you stills in the same call.