Free speech-to-text API credits compared, October 2026

Seven speech-to-text vendors publish a free allowance or starting credit on their pricing pages, from 60 minutes a month at Amazon to $200 at Deepgram. Sume's docs claim no free credit; STT is $0.01 per audio minute, so the dollar-denominated credits convert directly: $50 buys about 83 hours of Sume STT, $100 about 167 hours and $200 about 333 hours.

What each vendor lists

Each row is from that vendor's own pricing page, read 2026-10-02. The last column is our arithmetic at Sume's $0.01 per minute and ignores each vendor's own rates.

Using credits well

Spend credits on your own hardest audio, not a clean demo clip: crosstalk, accents, phone quality. Compare word-level timings too, since captions and dubbing depend on them. A Sume STT job returns words[] with start and end seconds.

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())

What free credit hides

A credit says nothing about the price after it runs out, and each vendor's paid rate differs. The comparison posts linked below put the paid rates next to Sume's.

Worked example

Here is a way to use several free allowances in one evaluation.

• Pick one 10-minute file with two speakers and background noise.
• Run it through each service's free allowance and compare names, numbers and timing.
• Then price a month of your real audio at each service's paid rate.

Checklist before you commit

At Sume's $0.01 per minute, a 10-minute test file costs $0.10, so even without a free credit a fair test is cheap.

• Read each terms page for credit expiry.
• Do not upload private audio to test accounts.
• Record the settings you used.