Transcribe a 25-minute interview on Sume: three STT jobs, $0.26

Sume STT 1.0 accepts 1 to 600 seconds of audio per job at $0.01 per audio minute, so a 25-minute interview needs chunks of 10, 10 and 5 minutes: three STT jobs at $0.25 plus a $0.01 timeline split job, $0.26 at list.

The steps

1. Host the file at a public HTTPS URL, since audio_url must be public.

2. Split with timeline audio into ranges (timeline audio docs): up to 20 ranges for a flat $0.01 per job.

3. Submit one STT job per range with an idempotency key, then add each chunk's start offset to its word times when you merge.

4. Omit language_code to auto-detect, or pass a BCP-47 hint.

Caveats

Words are always returned, with sentence segmentation optional. Speaker labels and audio-event tags are fixed server-side, not options. The docs say STT may appear on the dev API before the production OpenAPI lists it, so check GET /v1/catalog in your environment (API reference).

Sume STT is a batch job, not a live stream. If you need words while people are still talking, it is the wrong tool.