Paste a script and hear it read back in any of 10 voices, or upload a recording and get the words back with timestamps. Audio Studio plays a sample of every voice before you spend anything, and shows the price before you press Generate.
Turn a script into speech.
Pick a voice, paste up to 4,096 characters and press Generate. Ask for up to 4 takes, then download the one you like as MP3, WAV, FLAC, Opus and AAC.
Turn a recording into text.
Upload a file of up to 4 MB, or record up to 2:10 in the browser. With Timestamps on, you can download SRT subtitles as well as text.
curl --fail-with-body --retry 0 --max-time 120 \
https://api.minirouter.sh/v1/audio/speech \
-H "Authorization: Bearer $MINIROUTER_KEY" \
-H 'Content-Type: application/json' \
-D response.headers --output speech.mp3 \
--data '{
"model": "microsoft/mai-voice-2.1",
"input": "The quick brown fox jumps over the lazy dog.",
"voice": "en-US-Harper",
"response_format": "mp3"
}'Set MINIROUTER_KEY to your inference key.
See the price first.
Speech is priced per character and transcription per second, and the button shows the price before you send. Try it as a guest, no email needed.
- MAI-Voice-2.1
- $0.0231
- MAI-Voice-2.1-Flash
- $0.01575
- MAI-Voice-2-Flash
- $0.01575
- MAI-Voice-2
- $0.0231
- TTS-1
- $0.01575
- TTS-1 HD
- $0.0315
Estimated USD at current catalog rates.
Know the limits.
- Your history stays in this browser, up to your last 24 clips. Download the ones you want to keep.
- MAI-Voice speaks one voice, Harper, in US English. The OpenAI voices are multilingual.
- Studio uploads stop at 4 MB. For files up to 8 MiB, call
POST /v1/audio/transcriptionsdirectly; see the audio docs. - The acceptable use policy prohibits realistic audio of a real person saying something they did not, unless they agreed.