Voxtype vs VOXD

Two whisper.cpp-based tools with different philosophies: CLI daemon vs multi-UI application.

At a Glance

Aspect	Voxtype	VOXD
Engine	Whisper (whisper.cpp)	Whisper (whisper.cpp)
Language	Rust	Python
Interface	CLI + systemd	CLI, GUI, Tray
AI Post-Processing	No	Yes (local/cloud LLM)
Voice Activity Detection	No (push-to-talk)	Yes (FLUX mode, beta)
Distro Packages	AUR, .deb, .rpm	AUR, .deb, .rpm, pipx
GPU Required	No (optional)	No
GPU Acceleration	Vulkan, CUDA, Metal, ROCm	No
Text Processing	Word replacements, spoken punctuation	AI Post-Processing (LLM)

Critical Differences

Interface Philosophy

Voxtype is a headless CLI daemon. Configure it once, run it as a systemd service, interact via hotkey. No GUI, no tray icon needed.

VOXD offers multiple UI modes:

voxd --gui - PyQt6 graphical interface
voxd --tray - System tray icon for unobstructed dictation
voxd --rh - Terminal/CLI hotkey mode
voxd --flux - VAD-triggered continuous dictation

AI Post-Processing

Voxtype outputs exactly what Whisper transcribes. Clean, predictable, no surprises.

VOXD offers AI Post-Processing (AIPP) that can clean up transcriptions using a local or cloud LLM. This can fix grammar, add punctuation, or reformat text. Powerful but adds complexity and potential for unexpected changes.

Setup Experience

Voxtype setup is minimal:

paru -S voxtype
voxtype setup model
voxtype setup systemd
systemctl --user enable --now voxtype

VOXD has an interactive first-run setup:

# Install
paru -S voxd
# Or: sudo dpkg -i voxd_*.deb

# First run triggers setup wizard
voxd
# Configures: voice model, LLM model, ydotool

# For Wayland, also run:
./setup_ydotool.sh
# Then log out and back in

Feature Comparison

What VOXD Does Better

GUI options - Tray icon and graphical settings for users who prefer visual interfaces
AI post-processing - LLM can clean up transcriptions automatically
FLUX mode - Voice-triggered continuous dictation (no hotkey needed)
99+ languages - Explicit multilingual focus
Broader distro testing - Explicitly tested on Ubuntu, Fedora, Mint, Pop!_OS, openSUSE

What Voxtype Does Better

Simplicity - One binary, one config file, one systemd service
Predictable output - No AI surprises; what Whisper hears is what you get
Resource efficiency - Rust binary vs Python + PyQt6
GPU acceleration - Vulkan, CUDA, Metal, ROCm support for fast transcription
Text processing - Word replacements and spoken punctuation built-in
Waybar integration - Native status module for tiling WM users
Customizable audio feedback - Multiple themes, custom sounds

Resource Usage

Metric	Voxtype	VOXD
Idle Memory	~50MB	~100-150MB (GUI/tray)
Dependencies	Minimal (system libs)	Python, PyQt6, etc.
Binary Size	~15MB	Python scripts

The Verdict

Choose Voxtype if you prefer a minimal, headless daemon that runs via systemd. Ideal for tiling WM users who want Waybar integration and don't need a GUI.

Choose VOXD if you want a GUI or tray icon, need AI post-processing to clean up transcriptions, or prefer voice-triggered continuous dictation over push-to-talk.

Both are solid offline Whisper tools. VOXD offers more features and UI options; Voxtype offers simplicity and efficiency.