Voxtype vs WhisperWriter

Two Whisper-based dictation tools with different platform targets and workflows.

At a Glance

Platform Support

Voxtype

Linux-only, optimized for Wayland but works on X11 too. Uses evdev for hotkey detection (works on any Linux desktop), wtype for text output on Wayland (with CJK support), and ydotool on X11.

WhisperWriter

Cross-platform (Windows, Linux, macOS). Uses pynput for hotkeys and keyboard library for output. Wayland support has known issues.

Linux users: Voxtype has better Linux support due to its kernel-level evdev architecture. CJK character output works on Wayland via wtype.

Recording Modes

Voxtype: 2 modes

• Push-to-talk (default): Hold key to record, release to transcribe
• Toggle: Press to start, press again to stop

WhisperWriter: 4 modes

• Continuous (default): Records until silence
• Voice activity: VAD-based detection
• Press to toggle: Like Voxtype's toggle mode
• Hold to record: Like Voxtype's push-to-talk

API Options

Voxtype

Local Whisper only. Uses whisper.cpp for offline inference.

WhisperWriter

Can use either:

• Local Whisper model (faster-whisper)
• OpenAI API (requires API key)

The API option is useful if you have a weak CPU and an OpenAI account.

Setup Comparison

Voxtype

# Install
sudo dpkg -i voxtype_0.2.1-1_amd64.deb

# Interactive setup
voxtype setup model     # Select and download model
voxtype setup systemd   # Install as systemd service

WhisperWriter

git clone https://github.com/savbell/whisper-writer
cd whisper-writer
pip install -r requirements.txt
# May need CUDA setup for GPU
python run.py

Configuration

Voxtype: Config file

# ~/.config/voxtype/config.toml
[hotkey]
key = "rightctrl"
mode = "toggle"

[audio.feedback]
enabled = true

[text]
spoken_punctuation = true
replacements = { "javascript" = "JavaScript" }

WhisperWriter: GUI settings

PyQt5 settings window for configuration. Visual, but requires the app running.

Links

• Voxtype
• WhisperWriter on GitHub