Voxtype vs waystt

Two Rust-based speech-to-text tools for Wayland. Daemon vs signal-driven architecture.

At a Glance

Aspect	Voxtype	waystt
Engine	Whisper (whisper.cpp)	Whisper (cloud or local)
Language	Rust	Rust
Architecture	Systemd daemon	Signal-driven (starts/stops)
Default Mode	Offline (local)	Cloud (OpenAI API)
Hotkey Detection	Built-in (evdev)	External (WM keybinds)
Audio Feedback	Yes (customizable)	Yes (beeps)
Cursor Injection	Built-in (ydotool)	Via `--pipe-to`
GPU Acceleration	Vulkan, CUDA, Metal, ROCm	No
Text Processing	Word replacements, spoken punctuation	None

Critical Differences

Daemon vs Signal-Driven

Voxtype runs as a persistent systemd service. It's always listening for your hotkey, always ready. Start it once at login, forget about it.

waystt is invoked on-demand via signals. Your window manager keybind triggers it, it records, transcribes, outputs, then exits. This is lighter but requires more WM configuration.

Cloud vs Local Default

Voxtype is offline-only by design. Your voice data never leaves your machine.

waystt defaults to OpenAI's cloud API, requiring an API key and internet connection. Local mode with whisper-rs is available but requires downloading models separately.

Hotkey Handling

Voxtype handles hotkeys internally via evdev. Configure your key in config.toml and it just works.

waystt expects your window manager to send signals. You need to configure keybinds in Hyprland/Sway/etc. that run commands like pkill -SIGUSR1 waystt.

Setup Comparison

Voxtype

# Install and run - handles everything
paru -S voxtype
voxtype setup model
voxtype setup systemd
systemctl --user enable --now voxtype

# That's it. Hold ScrollLock to dictate.

waystt (Local Mode)

# Install
paru -S waystt-bin

# Download model
mkdir -p ~/.local/share/waystt
# Download GGML model manually...

# Configure WM keybinds (Hyprland example)
# bind = $mod, D, exec, waystt --local --pipe-to "ydotool type --file -"

# Start waystt in background, then trigger via keybind

Feature Comparison

What waystt Does Better

Minimal footprint - Not running when you're not using it
Cloud option - OpenAI API for those who prefer cloud accuracy
Google Speech support - Alternative cloud provider
Pipe-friendly - Outputs to stdout, easy to integrate with other tools

What Voxtype Does Better

Zero WM config - Hotkeys work out of the box, no keybind setup
Always ready - Daemon is instant, no startup latency
Privacy-first - No cloud option means no temptation to send voice data
GPU acceleration - Vulkan, CUDA, Metal, ROCm support for faster transcription
Text processing - Word replacements and spoken punctuation built-in
Integrated typing - Text appears at cursor automatically
Waybar module - Built-in status indicator

The Verdict

Choose Voxtype if you want a complete solution that handles hotkeys, transcription, and typing without additional WM configuration. Ideal for users who want "install and forget."

Choose waystt if you prefer a minimal, signal-driven tool that you invoke on-demand, or if you want cloud transcription options. Better for users who like composing tools via shell pipelines.