Voice KeyVOICE KEY / GUIDE
Get started with Voice Key
Install Voice Key on Windows, macOS or Linux, download a local speech model, and set up microphone access and voice input.
1. Install the desktop app
Download the latest release from the official GitHub repository. Choose the Windows x64 installer, the macOS disk image for Intel or Apple silicon, or a Linux x86_64 package. Current builds are unsigned, so Windows SmartScreen or macOS Gatekeeper may show a warning. Verify the source before opening the app.
2. Set up permissions and a microphone
In Settings, choose your microphone. On macOS, allow both Microphone and Accessibility access in System Settings → Privacy & Security. Accessibility permission enables global shortcuts and text insertion.
3. Download one recognition model
Classic mode uses SenseVoice (about 240 MB) and transcribes after recording. It recognizes Mandarin, English, Cantonese, Japanese and Korean. Streaming mode uses X-ASR (about 586 MB) and shows Chinese and English partial text in the HUD. Models are downloaded on demand and stored in your local application data directory. Only one mode is active at a time.
4. Try your first dictation
Focus a text field in your usual app. Hold the configured shortcut while speaking, then release to finish and insert the text. Defaults are Control+Shift+Space on Windows and Linux, and Option on macOS. Change the shortcut in Settings. Sessions can run for up to five minutes.
5. Choose whether to use AI
Basic local recognition needs no ASR API key. For optional refinement or translation, configure your preferred supported provider and credentials. These features send text to that provider and may incur provider charges. Leave both disabled for offline dictation after model setup.
Windows installation
Use the .exe installer for x64 Windows. If SmartScreen appears, first verify that the installer came from the official release page. The application is currently unsigned.
macOS installation
Select the .dmg matching your Mac architecture and move Voice Key to Applications. If Gatekeeper blocks the unsigned app, consult the macOS installation instructions in the repository. Enable Microphone and Accessibility permissions before trying voice input.
Linux and Omarchy
For Linux X11, use the x86_64 AppImage; make it executable before launching. Some distributions require FUSE 2. For Arch or Omarchy, use the pacman package. In Omarchy / Hyprland, open Settings → Linux & Omarchy Integration and install the managed integration. Other Wayland compositors do not yet have complete shortcut press/release support.
Need help?
Report reproducible issues through GitHub Issues. Include your operating system and recognition mode, but do not share credentials, private transcripts or raw recordings.
Useful links
Download the latest release · Project documentation & troubleshooting · GitHub Issues