Moonshine

Open-weights, low-latency speech-to-text models and a library for building voice agents and interfaces, with accuracy reported above Whisper Large v3.