Moonshine.
Transform speech-to-text with unmatched speed and precision
Experience industry-leading speech recognition with Moonshine: our family of fast, accurate, and lightweight open-weights models for edge devices. Our models are around 5X faster than OpenAI’s Whisper (scaled to the length of input audio) with better overall accuracy.
Key features.
Designed for the edge
Our models are built with edge devices in mind. They run directly on devices of all kinds—from laptops to Raspberry Pis—no internet connection or cloud provider required.
Lightning-Fast Processing
Moonshine models scale their processing requirements to the duration of input audio, empowering them to do more with less. Depending on the length of audio, they run anywhere from 10X to 1.5X faster than OpenAI’s tiny.en and base.en Whisper models.
Get started today!
Our open-weights models are permissively-licensed and support a wide variety of platforms. Try them right now in your browser with this demo link, check them out on GitHub to start building your own applications, or read our whitepaper to learn more.