v0.1.0

First release of TuxWhisper: offline, hotkey-driven dictation for Linux.

Features

• F5: tap or hold to dictate. Speech is transcribed locally with faster-whisper and pasted at the cursor.
• F3: dictate with cleanup through a local Ollama model (fixes punctuation and removes filler words).
• F4: save the last take and its transcript in Hugging Face audiofolder format.
• Works on X11 and Wayland, including GNOME on Wayland, in browsers and terminals.
• Restores your clipboard after each paste; pastes with Shift+Insert, so it works on any keyboard layout.
• NVIDIA GPU by default (large-v3-turbo); optional CPU mode.

Notes

• Tested on GNOME (Wayland). Reports from KDE, Sway, Hyprland and other desktops are welcome.

See the README for installation.