Local-first speech pipeline
Powered by Apple's SpeechAnalyzer and FoundationModels. Audio capture, transcription, and text cleanup all run on-device.
Native macOS App
Native macOS dictation that never leaves your machine.
Speak wraps Apple's on-device speech APIs in a fast command-loop: hold fn, dictate, release, and your text lands where you were already working.
Requires macOS 26+. Microphone and Accessibility permissions are required for full workflow.
Built for everyday dictation
Global hotkey control loop
Trigger from any app, stream partial results live, then commit into the focused text field in one motion.
Live volatile + final rendering
See rapid in-progress text and stable final segments so dictation feels responsive without being noisy.
Inspectable and scriptable
Open source by default. Clone it, build it, and evolve the workflow for your own team.
Under the hood
- SpeechAnalyzer — Apple's on-device speech-to-text engine (no Whisper, no cloud transcription)
- FoundationModels — Apple's on-device LLM for context-aware text cleanup
- Accessibility APIs — reads surrounding text for spelling and formatting context
Most dictation tools send audio to cloud APIs. Speak runs entirely on your Mac — no network required, no data leaves the device.
Will audio be uploaded?
No. Speak uses Apple's SpeechAnalyzer for transcription and FoundationModels for text processing — both run entirely on your Mac. No audio or text is sent to any server.
What if macOS blocks launch?
Open System Settings > Privacy & Security and click Open Anyway for Speak.
Can I build from source?
Yes. Clone, run make app, and ship your own fork.
Built for engineers, writers, and anyone who needs fast, private dictation across any app.
Read setup docs