Voice interfaces that never leave the device.
Five live demos of Moonshine Voice compiled to WebAssembly. The models download once, then run on your own CPU: no API key, no account, and not a single byte of audio sent to a server.
Speech to Text
Live transcription that keeps up with you as you talk, with the actual per-phrase latency shown next to every line.
Try the microphone →Text to Speech
Natural voices in 16 languages, plus a zero-shot clone of your own voice from about four seconds of speech.
Hear a voice →Voice Agent
A multi-step spoken conversation, triggered by meaning rather than exact wording. Say it however you like.
Start talking →Dictation App
A finished app rather than a demo: talk, and it types onto the page. Edit by hand, then take it away as Markdown or Word.
Start dictating →Meeting Notes App
Share a meeting and get a transcript that knows who spoke, keeps your edits attached to the audio, and plays any sentence back.
Take notes →