Repository navigation
Releases
VoxCommand for Windows 0.1.2
The Windows version of VoxCommand: say a phrase and it presses keys, clicks, types, launches apps, searches the web or talks back. Speech recognition runs locally with OpenAI's Whisper, so nothing leaves your PC.
Download and run
- Download
VoxCommand-0.1.2-windows-x64.zipbelow and unzip it anywhere (for exampleC:\Users\you\VoxCommand). - Run
VoxCommand.exe. The first start downloads the Whisper speech model (about 480 MB). - The program isn't code-signed, so SmartScreen may warn you the first time: choose More info > Run anyway.
voxcommand-cli.exe is the same program for the command line (voxcommand-cli.exe --help). Needs 64-bit Windows 10 or 11.
On Windows
- Keys and mouse clicks are sent as hardware scan codes, so games that read the keyboard directly (DirectInput) see them. To control a game or app that runs as administrator, run VoxCommand as administrator too.
- Push-to-talk works without extra permissions.
- Spoken replies use the voices built into Windows, or natural offline Piper voices (Settings > Spoken replies > Get voices…).
- Run application… lists your Start Menu programs. "Start VoxCommand when I log in" is in Settings.
- Optional NVIDIA GPU acceleration is an in-app download (Settings > GPU acceleration) and needs an NVIDIA driver.
- Your profile is stored in
%APPDATA%\VoxCommand\profile.yaml. Profiles work on both Linux and Windows.
Fixed in 0.1.2
- Sound actions beeped instead of playing the file on Windows. They now play wav, mp3, ogg, flac, m4a and other common formats. A new Sound file… button picks the file, and paths pasted with quotes work.
- New command
voxcommand-cli.exe play FILEto test a sound file.
New in 0.1.1 (both systems)
openaction: opens a web address, file, folder or shortcut with its default app. Add web search… now uses it.transcribeandsaycommands for the command line.
The Linux AppImage is in the Linux release.
VoxCommand Linux 0.1.2
VoxCommand Linux 0.1.2. Say a phrase and VoxCommand presses keys, clicks, types, launches apps, searches the web or talks back. Speech recognition runs locally with Whisper.
Download and run
chmod +x VoxCommand-0.1.2-x86_64.AppImage
./VoxCommand-0.1.2-x86_64.AppImageNew since 0.1.0
- Sound actions play wav, mp3, ogg, flac, m4a and other common formats. A Sound file… button picks the file, and a missing file is reported in the log.
openaction: opens a web address, file or folder with its default app, without a shell. Add web search… now uses it.- New commands:
transcribe FILEshows which command a recording would trigger,say TEXTspeaks with your voice settings, andplay FILEplays a sound file. - The same code now also builds for Windows (see the separate "VoxCommand for Windows" release). Profiles work on both.
See the README for permissions (/dev/uinput, the input group for push-to-talk) and the optional GPU download.
VoxCommand for Windows 0.1.1
The Windows version of VoxCommand: say a phrase and it presses keys, clicks, types, launches apps, searches the web or talks back. Speech recognition runs locally with OpenAI's Whisper, so nothing leaves your PC.
Download and run
- Download
VoxCommand-0.1.1-windows-x64.zipbelow and unzip it anywhere (for exampleC:\Users\you\VoxCommand). - Run
VoxCommand.exe. The first start downloads the Whisper speech model (about 480 MB). - The program isn't code-signed, so SmartScreen may warn you the first time: choose More info > Run anyway.
voxcommand-cli.exe is the same program for the command line (voxcommand-cli.exe --help). Needs 64-bit Windows 10 or 11.
On Windows
- Keys and mouse clicks are sent as hardware scan codes, so games that read the keyboard directly (DirectInput) see them. To control a game or app that runs as administrator, run VoxCommand as administrator too.
- Push-to-talk works without extra permissions.
- Spoken replies use the voices built into Windows, or natural offline Piper voices (Settings > Spoken replies > Get voices…).
- Run application… lists your Start Menu programs. "Start VoxCommand when I log in" is in Settings.
- Optional NVIDIA GPU acceleration is an in-app download (Settings > GPU acceleration) and needs an NVIDIA driver.
- Your profile is stored in
%APPDATA%\VoxCommand\profile.yaml. Profiles work on both Linux and Windows.
Also new in 0.1.1 (both systems)
openaction: opens a web address, file, folder or shortcut with its default app. Add web search… now uses it.transcribeandsaycommands for the command line.
The Linux AppImage is in the Linux release.
VoxCommand Linux 0.1.0
First release of VoxCommand Linux, an open-source alternative to VoiceAttack for Linux.
Say a phrase and VoxCommand presses keys, clicks, types, launches apps, searches the web or talks back. Speech recognition runs locally with OpenAI's Whisper (via faster-whisper), so nothing leaves your machine.
Download and run
chmod +x VoxCommand-0.1.0-x86_64.AppImage
./VoxCommand-0.1.0-x86_64.AppImageNo arguments opens the GUI; arguments use the command line (try --help).
Highlights
- Works on Wayland and X11, in native games and under Proton/Wine (input goes through the kernel's uinput)
- GUI command editor with an action guide, app picker, web-search builder and a test button
- Phrase alternatives (
[open|show] map), optional words and dictation (type *) - Actions: keys and combos, hold/release, typing, mouse, scroll, pauses, shell commands, spoken replies, sounds
- Spoken replies with system voices or natural offline Piper voices (download in-app or use your own
.onnx) - Wake word, push-to-talk, start at login, runs in the system tray
- Optional NVIDIA GPU acceleration as an in-app download (Settings > GPU acceleration)
Notes
- Built and tested on Fedora 45 (x86_64). Whisper models download on first use to
~/.cache/huggingface. - If key presses don't work, see "Permissions" in the README (
/dev/uinputaccess). Push-to-talk needs your user in theinputgroup.