Microsoft MAI-Transcribe-2-Streaming and MAI-Voice-2.1 Voice Models
Microsoft AI launched its first streaming transcription model, MAI-Transcribe-2-Streaming, on 1 October 2026, alongside two text-to-speech models, MAI-Voice-2.1 and MAI-Voice-2.1-Flash, all available through Microsoft Foundry.
Tool Overview
Features, steps and FAQ below
Features
- ✓ Low-latency real-time transcription in 60 languages with automatic, continuous language detection: Microsoft says it produces its first partials in just over 100ms of receiving audio
- ✓ Microsoft says it ranks no. 1 for accuracy for both final and partial transcripts on Artificial Analysis and sits on the Pareto frontier of its accuracy-versus-latency evaluation
- ✓ MAI-Voice-2.1 is Microsoft's strongest multilingual text-to-speech model, supporting 23 languages and 26 locales and letting a single voice speak each language with a native accent
- ✓ MAI-Voice-2.1-Flash targets high-volume, latency-sensitive workloads: Microsoft says it can generate 45s of audio with a 150ms end-to-end latency, is 55% faster at inference and about 60% cheaper
- ✓ Both voice models support voice cloning across all supported languages using just a few seconds of reference audio, with built-in consent guardrails to prevent misuse
How to Use
- Access the three models through Microsoft Foundry, the MAI Playground or Vercel, with the voice models also available via OpenRouter
- Call MAI-Transcribe-2-Streaming for streaming transcripts in voice agents, dictation or subtitling so apps can act before the speaker finishes
- Use MAI-Voice-2.1 to build multilingual customer-service agents or assistants that reply in the same voice with a native accent
- Choose MAI-Voice-2.1-Flash in high-volume scenarios for lower latency and cost with the same languages and cross-language voice
FAQ
What is MAI-Transcribe-2-Streaming?
Microsoft AI's first streaming transcription model, launched on 1 October 2026, which Microsoft says delivers low-latency real-time transcripts and produces its first partials in just over 100ms of receiving audio.
How many languages does it support?
Microsoft says MAI-Transcribe-2-Streaming supports 60 languages with automatic, continuous language detection.
How many languages does MAI-Voice-2.1 support?
Microsoft says MAI-Voice-2.1 supports 23 languages and 26 locales, letting one single voice speak across languages with a truly native accent.
How are these models priced?
Microsoft's blog notes an introductory price of $0.54 per hour of audio for MAI-Transcribe-2-Streaming through the end of the year, $22 per 1M characters for MAI-Voice-2.1 and $15 per 1M characters for MAI-Voice-2.1-Flash.
Where can I use these models?
Microsoft says all three are available through Microsoft Foundry and also via the MAI Playground and Vercel, with the voice models additionally available through OpenRouter.