What $100 buys in transcription hours: MAI Streaming vs Sume STT

$100 buys about 185 hours of audio on MAI-Transcribe-2-Streaming at Microsoft's $0.54 an hour introductory rate, and 166 hours and 40 minutes on Sume STT 1.0 at $0.01 per audio minute, which is $0.60 an hour. Microsoft says the $0.54 rate runs through the end of 2026.

Rates are from Microsoft's announcement and Sume's API reference and pricing page, read 2026-10-05. Hours are budget divided by rate; they ignore taxes and any minimum charge.

The conversion

A per-hour price and a per-minute price are the same unit once you multiply by 60.

What to add before you trust the number

The headline hours are a ceiling. Real recordings cost a little more on Sume because of how the work is shaped.

• Sume STT jobs are 1 to 600 seconds, so an hour of audio is six or more jobs; a split job at $0.01 flat adds about a cent per recording of up to 200 minutes.
• Send duration_seconds so a short clip reserves its real length: if it is omitted, Sume reserves one minute.
• Microsoft's intro rate ends on December 31, 2026; Microsoft's pages do not give the price after that date.

Check your own spend before you commit

This prints the hours a budget buys at each rate, so you can change the budget or the rate when Microsoft's price changes.

budget = 100.0
rates_per_hour = {"MAI-Transcribe-2-Streaming": 0.54, "Sume STT 1.0": 0.60}
for name, rate in rates_per_hour.items():
    hours = budget / rate
    print(f"{name}: {hours:.1f} hours ({hours * 60:,.0f} minutes)")