Soniox $0.10 per hour async vs Sume STT $0.01 per minute

Soniox's pricing page lists asynchronous transcription at $0.10 per hour and real-time at $0.12 per hour. Sume STT 1.0 has a public rate of $0.01 per audio minute, which is $0.60 per hour. On raw price per hour, Soniox is cheaper by a factor of six, and that is worth saying plainly before anything else.

The decision then comes down to what else the job needs: where the audio lives, how you call it, and what you do with the transcript afterward.

What each page says

Both numbers come from the vendors' own pages, read on 2026-10-02. Soniox bills by tokens and says the per-hour figures are an estimate of roughly 30,000 audio tokens plus 15,000 text tokens per hour, so your own audio may differ slightly.

What a Sume STT job looks like

You send a public HTTPS audio_url to POST /v1/stt-1.0/transcribe, optionally with language_code and a duration_seconds hint from 1 to 600. If you omit the hint, Sume reserves one minute. The result carries text and words[] with start and end times, and segmentation.mode set to sentence adds caption-shaped lines.

Because it is an ordinary Sume job, you poll status and result the same way as for video, image or music, and you can pass an Idempotency-Key so a retry does not double-submit.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/stt-1.0/transcribe",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
             "Idempotency-Key": "stt-demo-001"},
    json={"audio_url": "https://example.com/call.mp3",
          "language_code": "en",
          "duration_seconds": 300},
)
print(r.status_code, r.json())

When Sume's rate is still the right call

If transcription is one step in a pipeline that already runs on Sume, such as captions or dubbing, one key, one job model and one balance can be worth more than the per-hour difference. If you transcribe thousands of hours of long recordings and nothing else, price-per-hour dominates and a specialist is the honest answer.

Note the 600-second hint ceiling per request: long recordings need to be split into pieces first. The audio detach and 20-minute dubbing walkthrough shows how.

Check before you commit

Confirm the live Sume rate in GET /v1/catalog as the docs advise, and read Soniox's page again at purchase time, since both vendors can change rates.

Worked example

Take a backlog of recorded interviews. The arithmetic below uses only the two published rates and ignores any minimums, taxes or volume deals.

• 100 hours: Soniox async $10.00, Sume STT $60.00 (6,000 minutes at $0.01).
• 1,000 hours: Soniox $100.00, Sume $600.00.
• Sume's request cap is 600 seconds per job, so 100 hours means at least 600 jobs.

Checklist before you commit

Before you decide, run the same 10 files through both and compare errors on names, numbers and overlapping speech. Measure what you will actually pay for the whole path, including splitting, retries and storage of the transcripts.

• Check that the audio is reachable at a public HTTPS URL, since Sume fetches it.
• Decide whether you need word timings; Sume returns them, and sentence segmentation is optional.
• Re-read both pricing pages on the day you buy.