MAI-Transcribe-2
Microsoft AI speech-to-text model for 60 languages with diarization, word timestamps, keyword biasing and code switching.
Specs
- Context
- —
- Max output
- —
- Input price
- —
- Output price
- —
- Released
- 2026-09-03
Context
- Not published
Pricing
- Not published
Modalities
- Input
- Audio
- Output
- Text
API
- API types
Reasoning
- Not published
Info
- Status
- Preview
- Released
- 2026-09-03
- Knowledge cutoff
- —
Features
- Not published
How to call
1 provider
microsoftmodel = MAI-Transcribe-2
audio
POSThttps://<resource-name>.cognitiveservices.azure.com/speechtotext/transcriptions:transcribe?api-version=2025-10-15
JSON
Standard format
{
"id": "mai-transcribe-2",
"object": "model",
"created": 1788393600,
"owned_by": "microsoft",
"name": "MAI-Transcribe-2",
"api": {
"types": [
"audio"
]
},
"limits": {
"context": null,
"input": null,
"output": null
},
"modalities": {
"input": [
"audio"
],
"output": [
"text"
]
},
"reasoning": {
"supported": null,
"efforts": []
},
"pricing": null,
"features": [],
"info": {
"status": "preview",
"release_date": "2026-09-03",
"knowledge_cutoff": null,
"description": "Microsoft AI speech-to-text model for 60 languages with diarization, word timestamps, keyword biasing and code switching.",
"docs": "https://learn.microsoft.com/en-us/azure/ai-services/speech-service/mai-transcribe",
"verified_at": "2026-10-11"
}
}Official sources
2