Cheapest speech-to-text API per hour, October 2026

For plain file transcription, xAI's REST speech-to-text at $0.10 per hour is the lowest listed rate I found, and Sume at $0.60 per hour is the highest in this table. Sume's value is not the rate; it is word timings on every result, an HTTPS job interface and the same account as TTS, music and mixing.

Per-hour table

Hourly figures are the listed per-minute prices times 60.

Read the fine print

• OpenAI announced on 26 August that whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize shut down on 26 February 2027, so the gpt-4o-mini row is a short-lived option. See voice API deadlines.
• Gemini's price is two meters: audio in and text out. The combined figure assumes continuous speech.
• Sume reserves one minute if you omit duration_seconds; send the real duration (1 to 600 seconds) so the reservation matches the clip.
• Sume takes up to 10 minutes per request, so an hour is six or more jobs.

What the Sume result gives you

Every Sume transcription includes word-level timings, and you can request sentence segments with segmentation: {mode: sentence}. Speaker labelling and audio event tagging are fixed on the server side and are not options. Input is a public HTTPS audio URL; see the API reference.

Limits

Price per hour is not accuracy. None of these figures compares word error rate, and this post does not claim one. Test on your audio. If you need cheap bulk transcripts and no timings, a cheaper vendor wins. If you need the timings to cut video, Sume's cost is $0.01 per minute.