MAI-Voice-2.1
Microsoft AI's text-to-speech model covering 23 languages with a consistent voice identity, plus a low-latency Flash variant.
Overview
Microsoft AI released MAI-Voice-2.1 alongside MAI-Transcribe-2-Streaming on October 1, 2026. The model generates speech across 23 languages and 26 locales while keeping one recognizable voice consistent across all of them, avoiding the character shift that can occur in bilingual output. A companion MAI-Voice-2.1-Flash variant trades some quality for speed, generating 45 seconds of audio with about 150 milliseconds of lag for near-instant responses. Pricing is $22 per million characters.
This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.
Key features
- 23 languages across 26 locales with one consistent voice identity
- MAI-Voice-2.1-Flash variant: 45 seconds of audio generated with ~150ms lag
- Priced at $22 per million characters
- Released alongside Microsoft's first streaming STT model, MAI-Transcribe-2-Streaming
Use cases
- Multilingual voice assistants needing a stable voice identity across languages
- Low-latency voice responses for real-time conversational AI (Flash variant)
- Localized audio content and dubbing
Pricing
$22 per million characters.
Pricing reflects the source record and may change. Confirm current terms with the provider.
Dates and sources
- Released
- 1 Oct 2026
- Record updated
- 4 Oct 2026
- Record last checked
- 4 Oct 2026