MAI-Voice-1
In-house expressive speech model generates a minute of audio in under a second on a single GPU; powers Copilot Daily and Podcasts.
Microsoft's first self-built speech generator, single- and multi-speaker, deployed first in Copilot Daily, Copilot Podcasts and Copilot Labs; later sold via Foundry at $22 per 1M characters (April 2026).
- Date
- Thursday, 28 August 2025
- Lab
- Microsoft
- Kind
- model
- Access
- app only
- Price
- $22 per 1M characters (Foundry, 2026-04)
Figures
| Measure | Value | Measured by |
|---|---|---|
| Speed | 60 s of audio in <1 s on one GPU company claim | company |
Sources
- microsoft.ai/news/two-new-in-house-models/
- microsoft.ai/news/today-were-announcing-3-new-world-class-mai-models-available-in-foundry/
This record was checked against its sources on 6 October 2026. How we check