Release Notes
Changelog
Fully offline. Rebuilt speech recognition with SenseVoice — no cloud, no account, no internet required. First public open-source release.
- Offline ASR via SenseVoice — Chinese, English, Japanese, Korean, Cantonese
- Built-in punctuation and inverse text normalization
- Capsule floating window with real-time waveform visualization
- LLM polish with "Refining..." status display
- Usage statistics dashboard — 30-day charts, hourly distribution, recognition quality
- AI hotword extraction from dictation history
- Developer ID signed for Gatekeeper compatibility
- Open sourced under MIT license on GitHub
Rebuilt from scratch as a lightweight native Mac app. Faster, smaller, and works in every app including terminals and code editors.
- Text appears instantly at your cursor — no window switching, no copy-paste
- Optional AI cleanup to remove filler words and fix mishearing
- Custom vocabulary to improve accuracy for your own words and names
- All transcription history saved locally on your Mac
- Menu bar icon with quick access to history, vocabulary, and settings
- Remap a mouse button to Enter for quick confirmation
First working version. Proof of concept.