VoxType
An open-source voice-to-text tool designed for Linux and macOS, with native optimization for the Wayland display server protocol. It supports offline operation using local models and allows users to trigger voice transcription via customizable keyboard shortcuts, outputting the transcribed text directly to the current cursor position.
Timeline (last 90 days)
VoxType is an open-source speech-to-text tool designed specifically for Linux and macOS, with no Windows version
VoxType captures keystrokes by directly reading /dev/input event devices, bypassing Wayland's input isolation restrictions
After hands-on comparison testing, the blogger concluded that VoxType outperforms OpenWhisper on nearly every dimension, including fewer bugs, more intuitive configuration, more complete command-line tools, and a friendlier interface
VoxType's transcription process can run entirely on local models without requiring an internet connection
Arch Linux users can install VoxType via the AUR package voxtype-bin with the command yay -S voxtype-bin
Running voxtype-setup automatically creates the configuration directory and model directory and writes default settings, with the default shortcut key being Scroll Lock and the default mode being push-to-talk
After modifying VoxType's configuration, the service must be restarted via systemctl --user restart voxtype for the changes to take effect
VoxType uses Linux low-level event codes rather than key name strings to define shortcut keys, requiring the evtest tool to look up key code values, such as the Print key corresponding to code value 99
VoxType's default base.en model can only output English; using other languages requires switching to the multilingual base model and specifying the language option in the configuration
VoxType filters English filler words (such as "er" and "um") by default, but in German, "er" and "um" are content words and will be incorrectly swallowed, which must be resolved by fully enabling the [text] configuration block and redefining the filler_words list
2 more timeline events
All Facts (12)
VoxType is an open-source speech-to-text tool designed specifically for Linux and macOS, with no Windows version
50%UnverifiedVoxType captures keystrokes by directly reading /dev/input event devices, bypassing Wayland's input isolation restrictions
50%UnverifiedAfter hands-on comparison testing, the blogger concluded that VoxType outperforms OpenWhisper on nearly every dimension, including fewer bugs, more intuitive configuration, more complete command-line tools, and a friendlier interface
50%UnverifiedVoxType's transcription process can run entirely on local models without requiring an internet connection
50%UnverifiedArch Linux users can install VoxType via the AUR package voxtype-bin with the command yay -S voxtype-bin
50%UnverifiedRunning voxtype-setup automatically creates the configuration directory and model directory and writes default settings, with the default shortcut key being Scroll Lock and the default mode being push-to-talk
50%UnverifiedAfter modifying VoxType's configuration, the service must be restarted via systemctl --user restart voxtype for the changes to take effect
50%UnverifiedVoxType uses Linux low-level event codes rather than key name strings to define shortcut keys, requiring the evtest tool to look up key code values, such as the Print key corresponding to code value 99
50%UnverifiedVoxType's default base.en model can only output English; using other languages requires switching to the multilingual base model and specifying the language option in the configuration
50%UnverifiedVoxType filters English filler words (such as "er" and "um") by default, but in German, "er" and "um" are content words and will be incorrectly swallowed, which must be resolved by fully enabling the [text] configuration block and redefining the filler_words list
50%UnverifiedVoxType's base model provides a good balance between quality and resource usage, suitable for most scenarios, while tiny is suitable for lower-performance laptops
50%UnverifiedVoxType supports calling remote APIs (such as the OpenAI API) for higher transcription quality or to avoid occupying local hardware resources
50%