Speech to text
MAI-Transcribe 2
MAI-Transcribe 2 is a multilingual speech-to-text model from Microsoft AI, ranked #1 on the FLEURS multilingual benchmark. It supports 60 languages with automatic language identification, code switching for mixed-language speech, speaker diarization, word-level timestamps, keyword biasing for domain-specific terminology, and configurable verbatim or clean transcription styles. It is suited for captions, call transcription, subtitling, accessibility, and other voice-enabled applications, and is faster than MAI-Transcribe-1.5 on long-form audio.
WAV, FLAC, OGG, MP3 · Uploads up to 8 MiB
$0.0018Per minute
Price
- Per minute
- $0.0018
- Per hour
- $0.11
Includes the MiniRouter fee. Tokens are not charged.
Estimate your cost
Priced per second of audio.
Uploads up to 8 MiB.
Estimated cost
Includes the MiniRouter fee. Credit is held on an upper bound of the upload’s duration and released on settlement.
Estimate only; nothing is sent. Your charge settles to the provider’s reported seconds plus the MiniRouter fee.
Use the Transcription API
Set MINIROUTER_KEY to your inference key and copy the example.
Use a recording named recording.mp3, up to 8 MiB.
curl --fail-with-body --retry 0 --max-time 120 \
https://api.minirouter.sh/v1/audio/transcriptions \
-H "Authorization: Bearer $MINIROUTER_KEY" \
-F 'file=@recording.mp3' \
-F 'model=microsoft/mai-transcribe-2' \
-F 'response_format=json'Keep the request ID and check Request logs for the settled cost. A failed or disconnected request may still incur a cost; these examples disable automatic retries.
How it works
- Endpoint
- POST /v1/audio/transcriptions
- Multipart file and model, or JSON with base64 audio and media_type.
- Accepted media
- wav, flac, ogg, mp3
- MP4/M4A are rejected by this model.
- Response
- json, text, verbose_json
- verbose_json adds language, duration and segments.
- Billing
- Per second of audio
- Settles to the duration the provider reports.