Microsoft has released MAI Realtime 1, its first speech-to-speech voice AI model.
Unlike text-based models that convert speech to text and back, MAI Realtime 1 works directly with audio — listening and speaking in real time. It is designed for voice assistants, customer service and other applications where natural, low-latency conversation matters.
What changed
MAI Realtime 1 is Microsoft's first model built for speech-to-speech interaction. Key points:
- Speech-to-speech: the model takes audio in and produces audio out, without a separate text step.
- Low latency: designed for real-time conversation, with natural turn-taking.
- Pricing: starts at $0.01 per audio minute, according to Microsoft.
- Voice AI: part of Microsoft's broader MAI model family, announced at Microsoft Ignite 2026.
Metrics
| Metric | MAI Realtime 1 |
|---|---|
| Type | Speech-to-speech voice model |
| Starting price | $0.01 per audio minute |
| Use cases | Voice assistants, customer service, real-time conversation |
Why it matters
Voice is becoming a primary way people interact with AI, from phone calls to smart speakers. A model that handles speech directly — rather than through text — can sound more natural and respond faster. Pricing at $0.01 per audio minute makes it practical for high-volume voice applications.
Sources
- Microsoft AI — Introducing MAI Realtime 1
- Microsoft AI — Microsoft Ignite 2026
Originally published at iacompa.com
Top comments (0)