Local, privacy-first speech-to-text for macOS
Press a hotkey, speak, and text appears at your cursor.
All processing happens on-device — no audio or text ever
leaves your Mac.
Windows coming soon
Everything you need, nothing you don't
All audio capture and transcription happens on-device. No cloud APIs, no telemetry, no data collection. Your words stay on your Mac.
~44x real-time transcription on Apple Silicon. Powered by CoreML and the Apple Neural Engine for hardware-accelerated inference.
Press Cmd+Shift+Space to talk, release to transcribe. Or use Cmd+Shift+D to toggle hands-free recording. All shortcuts are fully customizable.
On-device LLM automatically removes filler words, fixes punctuation, and polishes your transcriptions — all locally with Qwen 3.5.
Multilingual speech recognition powered by NVIDIA Parakeet TDT v3 — a 600M parameter model supporting 25 languages out of the box.
Lives quietly in your menu bar. No dock icon, no distractions. Appears only when you need it with a beautiful floating overlay.
Browse, search, and copy past transcriptions. Every recording is saved locally so you never lose what you said.
Transcribed text is automatically pasted wherever your cursor is. Works in any app — editors, browsers, terminals, chat windows.
Just 14 MB app bundle. Models are downloaded once on first launch (~500 MB) and cached locally. Minimal resource usage when idle.
Get started in seconds
Built in the open, for everyone
SottoASR is fully open source under the MIT License. Inspect the code, contribute features, report bugs, or fork it and make it your own. Privacy you can verify — not just trust.
Built on the shoulders of giants
SottoASR relies on outstanding open-source libraries and models. All 660+ dependencies use permissive or weak-copyleft licenses (MIT, Apache-2.0, BSD, MPL-2.0, Unicode-3.0, ISC, Zlib, CC-BY-4.0). See THIRD_PARTY_LICENSES for the full list.
Author
Contributors
Built with significant assistance from Claude Code by Anthropic.
Want to contribute? Check out the GitHub repository.