Moonshine
An on-device AI voice toolkit for speech recognition, intent recognition and text to speech, with support for desktop, mobile, browsers and Raspberry Pi.
Moonshine is an on-device AI toolkit for developers building voice agents and applications that listen and speak. It combines speech to text, intent recognition and text to speech in one library. Voice processing stays on the device, and you don't need an account or API keys.
Its streaming approach processes speech while someone is still talking, reducing the delay before an application can respond. That makes it relevant to conversational interfaces where waiting for a complete recording would interrupt the exchange. Developers can also use it for transcription without building a full voice agent.
The speech recognition models are trained from scratch. The range includes models focused on accuracy and small models suited to constrained devices, so developers can choose according to the application and hardware. The documentation also covers model customization for particular domains.
Moonshine supports macOS, Windows and Linux, as well as iOS, Android and Raspberry Pi. Python and JavaScript/WASM support let developers use the toolkit across different application environments, including browsers. A C API is available for integration into native software.
The toolkit is open source under the MIT License. Its models use MIT by default across languages and sizes, with a specific exception: legacy non-streaming models for languages other than English use the non-commercial Moonshine Community License.
