Moonshine AI released a second generation of its open source voice toolkit, covering speech to text, text to speech and spoken conversational agents that run entirely on the device rather than in the cloud. The company reports its 245-million-parameter streaming English model transcribes with a lower error rate than OpenAI's much larger Whisper Large v3, and far faster. Packages are available for Python, JavaScript, iOS, Android, Windows, Linux and Raspberry Pi, with accuracy varying by language.
What changed
The first generation of Moonshine models topped out near Whisper Base accuracy, had no caching for streaming audio, and lacked a single cross-platform library.
What it unlocks
Running speech recognition, speech synthesis and simple spoken agents entirely on a phone, laptop, Raspberry Pi or microcontroller, without an account or API key.
- Moonshine Medium Streaming: 6.65% word error rate, 245 million parameters
- Whisper Large v3: 7.44% word error rate, 1.5 billion parameters
- 74ms vs 11,286ms transcription time on a MacBook Pro
- models down to 1MB for constrained devices
What you need to act on it
- installing the open source package for the chosen platform
- a microphone on the device
- github.com2026-08-05