Cost per minute of AI narration: MAI-Voice vs Sume at 900 characters

Text-to-speech is billed per character, but a budget is built in minutes of finished audio. To convert, this post assumes 150 spoken words a minute and 6 characters a word including the space, which is 900 characters per minute. That is an assumption, not a measurement of any vendor's voice; pace changes with the voice, language and punctuation.

Prices: Microsoft's MAI-Voice-2.1 is reported at $22 per million characters and MAI-Voice-2.1-Flash at $15 per million, launched 2026-10-01 (figures as reported by Digital Applied and PYMNTS, read 2026-10-06). Sume's TTS Router sells Cartesia Sonic at $0.038 list per 1,000 characters, so $0.0475 per 1,000 after the 1.25 margin.

One minute and one hour

What the gap is

On list price alone MAI-Voice-2.1 is $0.0198 a minute against Sume's $0.045101 all-in, a difference of about $0.025301 per minute of audio. Over 100 hours of narration a month that is $151.81. Sume does not host MAI voices today: its TTS Router lists Cartesia Sonic only, so the choice is between going direct to Microsoft and staying in one Sume job lifecycle with video, captions and timeline steps.

Sume prices every job at the vendor list price × 1.25, then adds a 5.5% fee on that result, so $1.00 of list becomes $1.31875 (how the formula works). The tables below show both the catalog price (list × 1.25) and the all-in figure with the fee.