Install on Linux
Speech Dock runs on Ubuntu, Debian, Fedora, Arch and openSUSE via .deb and AppImage. The Free tier requires no registration.
X11 or Wayland — it matters
On X11, auto-paste into the active window works right after installation. On Wayland, for security reasons, you need a separate service (ydotool) — a one-time setup. For a full breakdown of the differences and the workflow, see the voice input on Linux article.
Supported distributions
System requirements and tradeoffs.
Local processing depends on memory, disk and first-run setup. A GPU is optional, but supported accelerators may reduce processing delay.
Linux
- OSUbuntu 22.04+, Fedora 38+, Debian, Arch, openSUSE — best-effort
- Architecturex86_64
- RAM8 GB minimum · 16 GB recommended
- DiskThe required components download during first setup; keep several GB of free space
- GPUNot required; supported accelerators may reduce processing delay
- NetworkNeeded to download the required components, plus updates, checkout and license
- Integrationydotool (Wayland), xdotool (X11)
.deb — Debian / Ubuntu
Download the latest release and install via dpkg:
# Fetching the latest version…AppImage — Fedora / Arch / openSUSE
AppImage is a self-contained executable that does not require installation:
# Fetching the latest version…System dependencies
The launcher will offer to install missing packages on first run. The list depends on the distribution:
Debian / Ubuntu
sudo apt install libgtk-3-0 libsecret-1-0 libayatana-appindicator3-1 \libkeybinder-3.0-0 libnotify4 libstdc++6 libgomp1
Fedora / RHEL
sudo dnf install gtk3 libsecret libappindicator-gtk3 keybinder3 \libnotify libstdc++ libgomp
Arch / Manjaro
sudo pacman -S gtk3 libsecret libayatana-appindicator \libkeybinder3 libnotify
ydotool for Wayland (auto-paste)
On Wayland (GNOME, KDE Plasma, Sway, Hyprland), automatic paste into the active window requires the ydotoold daemon. On X11 it works out of the box via xdotool.
# Ubuntu/Debiansudo apt install ydotool# Fedorasudo dnf install ydotool# Archsudo pacman -S ydotool# Start the daemonsystemctl --user enable --now ydotoold# Add the user to the input group (for /dev/uinput)sudo usermod -aG input $USER# Log out and back in after this command
Troubleshooting
Speech Dock uses CUDA when:
nvidia-driver535+ is installed- A compatible GPU is present (compute capability 6.0+)
libnvidia-ml.sois available (ships with the driver)
If CUDA is not detected, the app falls back to CPU automatically. Check:
nvidia-smi # Should show the GPU and driver versionldconfig -p | grep nvidia-ml # Should find libnvidia-ml.so.1
GNOME Mutter doesn't implement zwlr_layer_shell_v1, so the overlay renders as a regular window (degraded UX). Sway / Hyprland / KDE Plasma 6.2+ provide full layer-shell overlay via libgtk-layer-shell0:
# Ubuntu/Debiansudo apt install libgtk-layer-shell0# Fedorasudo dnf install gtk-layer-shell# Archsudo pacman -S gtk-layer-shell
Verify the daemon is running and the user is in the input group:
systemctl --user status ydotooldgroups | grep input # should be in the list
Without the daemon, text is copied to the clipboard but not pasted automatically. Use Ctrl+V manually.
The launcher uses pkexec for elevation. On minimal installations polkit-gnome-authentication-agent-1 (or the KDE/XFCE equivalent) may be missing. Start it manually, or install dependencies via apt/dnf/pacman.
Make sure that:
- The file has execute permissions:
chmod +x ./SpeechDock-*.AppImage libfuse2is installed (Ubuntu 22.04+ may requiresudo apt install libfuse2)- Write access to
~/.local/share/is available for downloading models
Ready to try it? The Free tier requires no registration.