Release dates
— Microsoft launched MAI-Transcribe-2-Streaming for real-time transcription, along with the multilingual MAI-Voice-2.1 and its low-latency Flash variant. ↗
MAI-Voice-2.1-Flash
A low-latency multilingual text-to-speech model for real-time voice agents and high-throughput workloads, supporting 23 languages.
Introduced by Microsoft on 2026-10-01 as a separately named low-latency speech model; available through Microsoft Foundry, MAI Playground, OpenRouter, Vercel, and Azure Voice Live.
MODEL NUMBER049
- First public date
- 2026-10-01
- Organization
- Microsoft AI
- Family
- MAI Speech
01
Specifications
Only published specifications are shown.
02
Overview
Why it mattered
Microsoft released it as a separately named Flash model alongside MAI-Voice-2.1, adding a low-latency real-time branch to the speech line; this record does not infer separate architecture or weights.
03
Architecture & capabilities
Model capabilities
Audio generationMultilingualVoice cloningReal-time audio
04
Release history
Microsoft launched MAI-Transcribe-2-Streaming for real-time transcription, along with the multilingual MAI-Voice-2.1 and its low-latency Flash variant.
View source05
Related models
Select a card to view the related model.
06
Sources
01
Our first streaming transcription model debuts at no. 1 on Artificial AnalysisMicrosoft AI · Published 2026-10-01 · Checked 2026-10-01
Official · Verified02MAI-Voice-2.1 and MAI-Voice-2.1-FlashMicrosoft Learn · Checked 2026-10-01
Official · Verified03MAI-Transcribe-2-Streaming overviewMicrosoft Learn · Checked 2026-10-01
Official · VerifiedSuggest a correction
Copy this template, add the field, proposed value, and primary source, and share it with the archive maintainer.