gdictate

Free streaming voice dictation for Linux using Google Speech via Chrome's Web Speech API.

Press a hotkey, speak, release — text is pasted into any app.

How it works

Hotkey (evdev) → Python daemon ↔ WebSocket ↔ Chrome (hidden) → Google Speech API
                                                                      ↓
                                              Text pasted via wl-copy + ydotool

Chrome runs minimized and hidden from taskbar/Alt+Tab. It only serves as a bridge to Google's free speech recognition — no API keys needed.

Features

• Streaming transcription — see words as you speak (KDE OSD / notify-send on GNOME, Sway, Hyprland)
• System tray icon — colored status indicator + left-click to toggle recording
• Global hotkey — configurable key combo via evdev (default: Ctrl+Alt)
• Auto-paste — result pasted into focused window via clipboard
• Auto-setup — detects microphone, browser permissions, DE, and configures everything on first run
• Zero cost — uses Google Speech through Chrome, no API keys or accounts

Requirements

• OS: Linux (Arch, Fedora, Ubuntu/Debian)
• DE: KDE Plasma 6 recommended (full OSD support); GNOME/Sway/Hyprland work with notify-send fallback
• Display: Wayland
• Browser: Google Chrome or Chromium
• Python: 3.10+

Quick start

git clone https://github.com/bigidulka/gdictate.git
cd gdictate
./install.sh        # installs deps (Arch Linux)
python gdictate.py  # that's it

On first launch, Chrome opens automatically to request microphone permission. Click "Allow", close Chrome — it restarts hidden and you're ready to dictate. This only happens once.

Usage

python gdictate.py                          # default: Ctrl+Alt, ru-RU
python gdictate.py --key CTRL+SUPER         # custom hotkey
python gdictate.py --lang en-US             # English
python gdictate.py --test                   # 5s test, no hotkey
python gdictate.py --no-ui                  # terminal only, no tray/OSD
python gdictate.py --debug                  # show WebSocket messages
python gdictate.py --setup                  # force re-setup browser permissions

Hotkey

Combo	Description
Ctrl+Alt (default)	Press to start recording, press again to stop and paste
--key CTRL+SUPER	Use Super instead of Alt
--key CTRL+SHIFT	Use Shift instead of Alt

Languages

Default is ru-RU. Pass --lang with any BCP-47 language tag:

python gdictate.py --lang en-US    # English
python gdictate.py --lang de-DE    # German
python gdictate.py --lang uk-UA    # Ukrainian

Compatibility

System	Support
Arch + KDE Plasma 6 + Wayland	Tested, full OSD + tray
Fedora / Ubuntu + KDE + Wayland	./install.sh supported, full OSD + tray
GNOME / Sway / Hyprland + Wayland	Works — tray icon + notify-send for streaming & final text
X11	Not supported (uses wl-copy, ydotool on Wayland)
macOS / Windows	Not supported

Architecture

┌─────────────┐     ┌──────────────────────┐     ┌─────────────────┐
│  evdev      │     │  Python daemon       │     │  Chrome (hidden) │
│  hotkey     │────>│  aiohttp WS server   │<───>│  Web Speech API  │
│  listener   │     │  state machine       │     │  (speech-proxy)  │
└─────────────┘     └──────┬───────────────┘     └────────┬────────┘
                           │                              │
                    ┌──────┴──────┐               ┌───────┴────────┐
                    │ wl-copy +   │               │ Google Speech  │
                    │ ydotool     │               │ servers (free) │
                    │ (paste)     │               └────────────────┘
                    └─────────────┘

Files:

• gdictate.py — main daemon: WebSocket server, Chrome launcher, hotkey listener, paste
• overlay.py — OSD overlay (KDE native / notify-send fallback) + system tray icon with click-to-toggle
• speech-proxy.html — Chrome page running webkitSpeechRecognition

License

MIT