Microsoft AI Releases MAI-Transcribe-2-Streaming: #1 Real-Time Speech-to-Text Model on Artificial Analysis
Microsoft AI has released MAI-Transcribe-2-Streaming, its first real-time speech-to-text model. It ranks #1 of 38 models on Artificial Analysis AA-WER Streaming. It scores 2.5% WER at 0.13s on final transcripts and 2.5% at 0.12s on first partials.
Microsoft AI has launched MAI-Transcribe-2-Streaming, a real-time speech-to-text model that ranks #1 in accuracy on the Artificial Analysis AA-WER Streaming index. It offers high accuracy with low latency, making it ideal for applications like voice agents and live captions.
What happened
Microsoft AI released MAI-Transcribe-2-Streaming, a real-time speech-to-text model that outperforms other models in accuracy and speed. It is part of a suite of AI tools, including MAI-Voice-2.1 and MAI-Voice-2.1-Flash, and is available through multiple integration paths.
Why it matters
The model's high accuracy and low latency make it ideal for applications where real-time transcription is critical, such as voice agents and live captions. Its performance on the AA-WER Streaming index highlights its competitive edge in the market.
- MAI-Transcribe-2-Streaming is the top-performing model on the AA-WER Streaming index with 2.5% WER.
- It provides accurate first partial transcripts within 0.12 seconds.
- Supports 60 languages with continuous language detection.
- Pricing is higher than xAI and Meta but competitive with Google's estimates.
- Available through multiple integration paths for developers.
The bigger picture
MAI-Transcribe-2-Streaming can significantly improve user experiences in voice-based applications by enabling faster and more accurate transcription. Its integration with MAI-Voice-2.1-Flash allows for seamless voice loops, enhancing the capabilities of voice assistants and other real-time communication tools.
MAI-Transcribe-2-Streaming represents a significant advancement in real-time speech-to-text technology, offering high accuracy and low latency. Its performance and integration options make it a valuable tool for developers working on voice-based applications.
Technical details
- Supports 60 languages with continuous automatic language detection.
- Emits first hypotheses (partials) within 100ms of receiving audio.
- Uses SileroVAD to detect the end of speech for latency measurement.
- Available via Realtime API and Azure Speech SDK for integration.
- Pricing at $0.54 per hour for an introductory period.
FAQ
What is MAI-Transcribe-2-Streaming?
MAI-Transcribe-2-Streaming is Microsoft's first real-time speech-to-text model that ranks #1 in accuracy on the Artificial Analysis AA-WER Streaming index.
What is the accuracy of MAI-Transcribe-2-Streaming?
It achieves 2.5% WER at 0.13 seconds for final transcripts and 2.5% at 0.12 seconds for first partials.
How many languages does it support?
It supports 60 languages with continuous automatic language detection.
What is the pricing for MAI-Transcribe-2-Streaming?
It costs $0.54 per hour for an introductory period through the end of 2026.
How can developers integrate MAI-Transcribe-2-Streaming?
Developers can integrate it via the Realtime API or Azure Speech SDK, which handles connection management and audio streaming.
This summary and analysis were generated with AI from the original article and may contain errors. Read the original for the full story.
Read the full story
Originally published by MarkTechPost on Oct 3, 2026.
Comments
Sign in to join the discussion.