MAI-Voice-2.1

Overview

MAI-Voice-2.1 is Microsoft‘s expressive text-to-speech model, released on 2026-10-02 — 23 languages with native-accent synthesis; a MAI-Voice-2.1-Flash variant hits 150ms latency; voice cloning from a few seconds of audio with safeguards (~half of 4,000 testers perceived cloned voices as real). Pricing: $22/1M chars for Voice-2.1, $15/1M chars for Flash. Shipped externally through Microsoft Foundry in public preview (no SLA). Framed by the corpus as the second iteration of Microsoft’s post-renegotiation OpenAI-independence voice stack — paired with MAI-Transcribe-2-Streaming, additive to Azure OpenAI rather than a mandated swap.

Timeline

• 2026-10-03-AI-Digest — MAI-Voice-2.1 launches on 2026-10-02 across 23 languages with native-accent synthesis, with a 150ms-latency Flash variant and voice cloning from a few seconds of audio. Testing disclosure: ~half of 4,000 testers perceived cloned voices as real. Intro pricing $22/1M chars for Voice-2.1 and $15/1M chars for Flash. Corpus framing: second iteration of Microsoft's post-renegotiation independence stack, not Microsoft fires OpenAI's voice stack — strategic turn was the 2025 contract renegotiation, today’s release is the hardening pass. Shipped in public preview on Microsoft Foundry — no SLA, side-by-side with Azure OpenAI, not a replacement.

Key Developments

1. MAI-Voice-2.1 and MAI-Voice-2.1-Flash public preview on Foundry (October 2, 2026): Expressive TTS across 23 languages with native-accent synthesis; Flash variant at 150ms latency; voice cloning from seconds-level reference audio with safeguards (~half of 4,000 testers perceived cloned voices as real). Pricing: $22/1M chars (Voice-2.1), $15/1M chars (Flash). Launched alongside MAI-Transcribe-2-Streaming. Public preview — no SLA, additive to Azure OpenAI on Foundry.

Related

See also: Microsoft, MAI-Transcribe-2-Streaming, MOC - Major Companies, MOC - Developer Tools.