SabreVoice is a free, open-source dictation application for Windows, Linux, and macOS. Dictate directly into any text input with local offline AI model processing, global hotkeys, and zero data leaving your computer.
git clone https://github.com/sabreproducts/SabreVoice.git
Also supports Linux via native Python install • GPLv3 License
Local speech-to-text models process all audio fully on-device. Perfect for dictating without internet connection or in secure local environments.
A gorgeous, draggable Tkinter HUD overlay showing active target application detection, live bouncing equalizer bars, and layout control.
Built-in silence/pause detection automatically transcribes and types your text in chunks as you speak, keeping your flow uninterrupted.
Format and rewrite your text into emails, clean up spelling and grammar, or summarize into bullet points using local or cloud LLMs.
| Model / Engine | Type | Size | Languages | Target Hardware |
|---|---|---|---|---|
| Moonshine Tiny | Local (Offline) | 33 MB | English only | Low-resource Windows & Linux CPUs |
| Moonshine Streaming Medium | Local (Offline) | 282 MB | English only | Standard office PCs & Laptops |
| Nemotron Speech 3.5 | Local (Offline) | 670 MB | 40+ Languages | Apple Silicon Macbooks (m1/m2/m3/m4) |
| OpenAI / Groq Whisper | Cloud (Online API) | None (API) | 99+ Languages | Active internet connection (API Key) |
Standalone background tray app with offline Moonshine support and visual HUD overlay.
Download sabre_voice.exeRebranded native SwiftUI macOS client supporting local Apple Speech, Nemotron and Whisper.
Download macOS DMGPython-based setup for Ubuntu, Debian, and Mint with simple installation script.
Get setup_linux.sh