Repolode
Submit
moonshine-ai

moonshine-ai/moonshine

Open source on-device toolkit for speech-to-text, intent recognition and text-to-speech, aimed at real-time voice agents and apps.

Open-source alternative to:Whisper

View on GitHubWebsite
Stars
11,260
Forks
620
Open issues
12
Last commit
18 days ago

What is it?

Moonshine Voice is an open source AI toolkit for building real-time voice agents and applications. It runs entirely on-device, so it needs no account or API keys and keeps audio local.

The speech-to-text models are trained from scratch and range from small models around 1MB for constrained devices up to larger models aimed at higher accuracy. The library is designed for live streaming, doing work while the user is still speaking to reduce response latency. It provides one API across Python, JavaScript/WASM, iOS, Android, macOS, Linux, Windows and Raspberry Pi.

The project also covers intent recognition and text-to-speech, and includes a coding-agent skill that teaches assistants the current API. Models are MIT licensed by default, with a few legacy non-English models under a non-commercial license.

Key features

  • 01Speech-to-text models trained from scratch, from very small to higher-accuracy sizes.
  • 02Live streaming transcription that starts processing while the user is still talking.
  • 03Runs fully on-device with no account or API keys required.
  • 04Single library across Python, JavaScript/WASM, iOS, Android, desktop and Raspberry Pi.
  • 05Includes intent recognition and text-to-speech for voice interfaces.

Use cases

  • Adding voice commands to a mobile or desktop app without sending audio to a cloud service.
  • Building a voice agent that responds quickly while the user is still speaking.
  • Running speech recognition on small embedded devices such as Raspberry Pi.

Quick start

  1. 1

    Install the Python package

    pip install moonshine-voice
  2. 2

    Run live microphone transcription in English

    moonshine-voice mic --language en
  3. 3

    For other platforms, follow the Quickstart guide

Things to know

  • Legacy non-streaming models for languages other than English use a non-commercial license.

Topics

#speech-to-text#text-to-speech#voice-agent#on-device#real-time#speech-recognition#edge-ai#intent-recognition#stt#tts

FAQ

What is moonshine?
Open source on-device toolkit for speech-to-text, intent recognition and text-to-speech, aimed at real-time voice agents and apps. Moonshine Voice is an open source AI toolkit for building real-time voice agents and applications. It runs entirely on-device, so it needs no account or API keys and keeps audio local.
Is moonshine free to use? What license does it have?
moonshine is open source under the Other license. Check the license terms before using it in commercial products.
Is moonshine still maintained?
Actively maintained. The last commit to moonshine was on Oct 2, 2026.
What can I use moonshine for?
Adding voice commands to a mobile or desktop app without sending audio to a cloud service. Building a voice agent that responds quickly while the user is still speaking. Running speech recognition on small embedded devices such as Raspberry Pi.
How do I install moonshine?
Install the Python package: pip install moonshine-voice Run live microphone transcription in English: moonshine-voice mic --language en For other platforms, follow the Quickstart guide
Where can I get moonshine?
The source code and installation instructions are on GitHub: https://github.com/moonshine-ai/moonshine

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects