← Back to explorer

MAI-Voice-2.1

Type
model
Venue
Microsoft AI announcement, Oct 1, 2026
Year
2026
Source
blog
Access
paid
Language
English
Added
2026-10-01
Verified
2026-10-01

Summary

Strongest multilingual text-to-speech model: 23 languages and 26 locales with a single cross-language voice. $22 per 1M characters.

Keywords

text-to-speech · TTS · voice agents · multilingual · cross-language voice

Topics

speech, text-to-speech, voice agents

Research notes

  • Discovery: @MicrosoftAI X post 2026-10-01 (https://x.com/MicrosoftAI/status/2105693024013467905)
  • Blog: https://microsoft.ai/news/our-first-streaming-transcription-model/
  • Demo: Chatter live voice-agent demo at http://playground.microsoft.ai
  • Access: OpenRouter (https://aka.ms/mai-openrouter), Microsoft Foundry, Vercel (https://aka.ms/mai-vercel), LiveKit (coming soon), Azure Voice Live
  • No Hugging Face/arXiv/GitHub release listed