Microsoft AI drops transcription price to 10 cents per hour
MAI-Transcribe-2 pairs speed, accuracy, and multilingual support with a limited introductory price.
Microsoft AI transcription: what changed with MAI-Transcribe-2
Microsoft AI transcription took a notable step forward with the announcement of MAI-Transcribe-2, an audio transcription model that combines speed, accuracy, and a very low introductory price. The core idea is straightforward: turn large amounts of audio into text quickly, with multilingual support and workflow features aimed at business use.
What changed
The main shift is not only technical but operational. The model is said to transcribe one hour of audio in ten seconds and launched at 10 cents per audio hour. For teams handling meetings, calls, or interviews, that can shorten the time between capturing information and putting it to work.
Why it matters for business
When transcription becomes faster, documentation, search, and summarization can also move faster. In audio-heavy environments, that may reduce manual effort and improve traceability. Still, the real value depends on audio quality, language mix, vocabulary, and how much human review your process requires.
Microsoft AI transcription: capabilities worth evaluating
MAI-Transcribe-2 ranks first in FLEURS, a public multilingual benchmark, with an average word error rate of 5.2% across 60 languages. Microsoft also cites third-party testing by Artificial Analysis showing it is faster than OpenAI's GPT-Transcribe, ElevenLabs' Scribe v2, and Gemini 3.5 Transcribe. These comparisons help frame performance, but they do not replace testing on your own audio.
Relevant features
- Speaker diarization to separate who said what.
- Word-level timestamps for precise editing and navigation.
- Keyword biasing for terminology, abbreviations, and proper nouns.
- A verbatim output for compliance and a clean output for publishing.
- Automatic language identification.
- Code switching support, useful when a conversation mixes Spanish and English.
Business scenarios
Microsoft has identified initial use cases in clinical notes, legal documentation, accessibility, and subtitling. In accessibility and subtitles, automation can speed up audiovisual publishing. In clinical notes and legal documentation, the operational upside is real, but human review remains essential to reduce error risk.
Limitations, risks, and caution flags
There are three important limits. First, the 10-cent price is a limited-time offer valid until the end of the year; the regular price has not been announced. Second, although it can already be tried through Microsoft Foundry, MAI Playground, and OpenRouter, general availability has not been confirmed and these are currently demos. Third, in critical use cases, automatic transcription does not remove the need for human oversight.
Evaluation checklist
- Test with real company audio.
- Check performance with noise, accents, and mixed languages.
- Measure the usefulness of diarization and timestamps.
- Validate company-specific terms with keyword biasing.
- Define where human review is mandatory.
How to apply it in your business
Start with a focused pilot of Microsoft AI transcription in a high-volume workflow such as meetings, support calls, or subtitling. Compare time saved, text quality, and review effort. If the model fits your audio and risk profile, it can become a useful layer for scaling documentation and content access.