Skip to content

Voxtype

Voxtype is a push-to-talk voice dictation tool. Hold a key, speak, release — the transcription is typed into whatever is focused. Built on OpenAI Whisper, compiled with Vulkan GPU acceleration for real-time inference on FreeBSD.

voxtype-install

Builds from source using cargo with the gpu-vulkan feature flag. Requires wtype and pipewire. Downloads the base.en Whisper model automatically on first install.

ActionCommand
Start recordingvoxtype-start
Stop and transcribevoxtype-stop
Check statusvoxtype-status

The keybinding Super + Z is configured as hold-to-record: press to start, release to transcribe and type the result.

ShortcutAction
Super + Z (hold)Record audio — release to transcribe and type

If vulkaninfo is available at install time, Voxtype is configured to use Vulkan for Whisper inference. This keeps transcription fast without requiring CUDA or a Linux environment.

voxtype setup gpu --enable # enable Vulkan manually
voxtype status # confirm GPU is active

Default config is installed to ~/.config/voxtype/config.toml. Edit to change model, language, or output method.