MAI-Voice-2.1 vs Flash: what $7 per million saves on a 60-second short

Microsoft listed MAI-Voice-2.1 at $22 per million characters and MAI-Voice-2.1-Flash at $15 on 2026-10-01 (Digital Applied tracker, read 2026-10-06). The tracker repeats Microsoft's claims that Flash is 55% faster and 60% cheaper; against the $22 sibling the list price is 31.8% lower, so the 60% presumably compares against something else, and the source does not say what. Use the list prices.

For narration, the saving is $7 per million characters. On a single 750-character, roughly 60-second short, that is $0.00525. It is a bulk-volume number, not a per-clip one.

Savings by volume

What the table cannot say

Price tells you nothing about sound. Flash is described as the faster, cheaper tier; whether it is good enough for your voiceover is a listening test. Both Microsoft models, as listed, support 23 languages and 26 locales (PYMNTS).

On Sume, the choice of model for narration is a TTS Router id; Sume TTS 1.0 lists $0.0475 per 1,000 characters, which is $0.03562 for the same 750 characters, rounded up to $0.04 per job.

A decision rule

• Under about 100,000 characters a month, the $7 per million saving is under a dollar. Choose on sound and language.
• Above tens of millions, the saving is real money; run a blind listening test first.
• Always price per job after your own vendor's rounding rules, not just per-character list prices.