Microsoft ships MAI-Voice-2.1 and Flash
MAI-Voice-2.1 is a text-to-speech model covering 23 languages where a single voice keeps its identity across languages. The Flash variant targets high-volume use with about 150 ms end-to-end latency.
What it does
MAI-Voice-2.1 is Microsoft's text-to-speech model for 23 languages and 26 locales. One voice can switch languages and keep the same identity with a native accent. MAI-Voice-2.1-Flash supports the same languages for high-volume, low-latency use.
Both can clone a voice from a few seconds of reference audio, with built-in consent guardrails.
Results and limits
Microsoft says Flash generates 45 seconds of audio with 150 ms end-to-end latency, with 55% faster inference and about 60% lower cost than comparable models. These are Microsoft's own figures, and the source gives no independent tests.
Worth knowing
Pricing is $22 per 1M characters for MAI-Voice-2.1 and $15 per 1M for Flash. Both are available via OpenRouter, Microsoft Foundry, MAI Playground and Vercel.
Try it yourself
In MAI Playground, enter: "[Narration line in English]" then "[same line in German]", using one voice for both. Listen for whether the speaker stays the same across languages.