MAI-Transcribe 1.5 transcription online

What is MAI-Transcribe 1.5?

MAI-Transcribe 1.5 is Microsoft's latest automatic speech recognition model, currently ranked third on the OpenASR leaderboard with a 2.4% word error rate. It is designed for production workloads that need high accuracy, low latency, and the ability to guide the model towards domain-specific vocabulary before a transcription run begins.

The key feature that sets it apart from general-purpose ASR models is entity biasing: you can supply up to 200 words or phrases — names, product terms, medical terminology, legal concepts — that the model should expect to encounter. The decoder is biased towards these candidates when audio is phonetically ambiguous, which substantially reduces first-pass errors on proper nouns and specialist vocabulary.

Speed is also a distinguishing characteristic. MAI-Transcribe 1.5 processes audio at under 15 seconds per hour of recording, compared to 60–90 seconds per hour for Whisper. For a 60-minute interview, the practical difference is receiving your transcript in under 8 seconds rather than waiting a minute or more.

MAI-Transcribe 1.5 vs. Whisper

Both models cost €0.02/min. Full comparison →

Frequently asked questions

What is MAI-Transcribe 1.5?

MAI-Transcribe 1.5 is Microsoft's latest automatic speech recognition model, currently ranked third globally on the OpenASR leaderboard with a 2.4% word error rate. It is designed for high-accuracy, low-latency transcription and supports up to 200 entity-biasing phrases — proper nouns, domain terms, or names that should be transcribed exactly. convert.express offers it as a premium model option alongside OpenAI Whisper.

How accurate is MAI-Transcribe 1.5 compared to Whisper?

MAI-Transcribe 1.5 achieves roughly 2.4% word error rate on benchmark datasets, compared to approximately 5–8% for Whisper on equivalent material. The practical difference is most visible on proper nouns, entity-heavy content (names, places, organisations), and material where entity biasing has been pre-loaded. On low-resource or niche languages, Whisper may be more reliable since it covers 57 languages vs. 43 for MAI.

What is entity biasing and how do I use it?

Entity biasing (also called vocabulary hints) lets you supply a list of words and phrases — up to 200 — that the model should expect to encounter. This primes the ASR decoder to prefer these candidates when the audio is ambiguous. Use it for: interviewee names, organisation names, product names, technical terminology, or any proper noun that a general-purpose model might mis-transcribe. Enter the phrases in the transcription settings before uploading your file.

How fast is MAI-Transcribe 1.5?

MAI-Transcribe 1.5 processes audio at under 15 seconds per hour — substantially faster than Whisper's 60–90 seconds per hour. A 60-minute podcast episode typically finishes in 3–8 seconds. This makes it practical for time-sensitive workflows where waiting a minute for a transcript is not acceptable.

What does "Public Preview" mean?

MAI-Transcribe 1.5 is in public preview on convert.express, which means it is available but carries no uptime SLA. In rare cases of high load or model unavailability, the service may fall back to OpenAI Whisper automatically. If your workflow depends on guaranteed availability, choose Whisper as the explicit model. Preview status also means the model may be updated at any time, which could change output format or accuracy characteristics.

When should I choose MAI-Transcribe 1.5 over Whisper?

Choose MAI-Transcribe 1.5 when: speed is important (interviews, live-event transcription); accuracy on proper nouns matters (journalism, legal, medical); you can provide entity biasing phrases to further improve accuracy; and the language you need is in its 43-language set. Choose Whisper when you need a language outside MAI's coverage, need English translation of a foreign-language recording, or prefer a generally-available model with formal SLA guarantees.