60 scripts of 800 characters: MAI-Voice-2.1 $1.06 vs Sume $2.40

The short answer

A third-party release tracker lists Microsoft's MAI-Voice-2.1 text-to-speech at $22 per million characters, and MAI-Voice-2.1-Flash at $15 per million. Sume does not list either model. Sume's own TTS Router bills $47.50 per million characters ($0.000038 list x 1.25 = $0.0000475 per character). On those numbers a 1,000-character script is about 2.2 cents at the MAI list, 1.5 cents on MAI Flash and 5 cents on Sume.

Treat the MAI figures as unverified. They come from the Digital Applied tracker (read 2026-10-08), not from a Microsoft price page, and the tracker describes MAI-Voice-2.1 as supporting 23 languages and the Flash variant as having a 150 ms latency claim.

Per-script arithmetic

Microsoft rows are the tracker's per-million rates divided by the character count. The Sume row is the billable rate, rounded up to a whole cent per job.

What the comparison leaves out

Price per character is a weak basis on its own. The Sume rate covers the Sonic engines listed in the TTS Router (five ids: sonic-3.6, sonic-3.5, sonic-3, sonic-latest, sonic-preview) with a documented 20,000-character request limit. The tracker does not state how MAI bills rounding, minimums, voice cloning or commercial terms. Voice quality, language coverage and the legal terms for use in ads are not captured by a single number.

Sume's rate also includes the 1.25 factor on top of provider list. The provider list is $38 per million characters, so the pure list-to-list comparison is $22 (MAI, per the tracker) against $38 (Sonic list). The 1.25 factor takes Sume to $47.50. Both views are legitimate; the second is what you pay on Sume.

How to decide

If your only goal is the lowest cost per character and MAI is available to you directly, the tracker's numbers suggest it is cheaper per character, subject to confirming the rate with Microsoft. If you want narration inside the same wallet, job format and timeline as your video and music, price it from the Sume catalog. The route map is in the API reference, and GET /v1/tts-router/models gives the live price.

For a 60-script batch of 800 characters each, MAI at $22/M would be 60 x 800 x $0.000022 = $1.056, and Sume would be 60 x 4 cents = $2.40 (800 characters is 3.8 cents, rounded up to 4). The money is small either way; the workflow is usually the deciding factor.