Local, offline voice-to-text for Linux — powered by Whisper Large V3.

Press a hotkey, speak, release. Your words appear as text instantly — no cloud, no latency, no subscription.

• Fully offline — Whisper runs locally on your GPU. Nothing leaves your machine.
• Global hotkey — Hold Ctrl+Super+Space to talk, release to transcribe. Or toggle mode.
• 20+ languages — Auto-detect or lock to Vietnamese, English, Japanese, and more.
• Dark & Light themes — Switch instantly in Settings → Appearance.
• Real-time waveform — Live audio visualization while recording.
• Transcript history — Search, edit, tag, and export all your transcriptions.
• Audio ducking — Automatically mutes other apps while you record.
• Sound feedback — Chirps on record start/stop, chime on transcript ready.
• Floating overlay — Compact always-on-top indicator during recording.
• Smart text injection — Paste, type (xdotool), or copy-to-clipboard output modes.
• Dashboard & gamification — Track usage stats and earn milestones.
• Session analysis — AI-powered insights on your transcription patterns.

Download the latest .deb from Releases, then:

sudo dpkg -i voicetotex_1.1.0_amd64.deb

git clone https://github.com/riftzen-bit/voicetotex.git
cd voicetotex
bash scripts/setup.sh
bash scripts/start.sh

python3 -m venv venv && source venv/bin/activate
pip install -r backend/requirements.txt
npm install
npm start

Output modes: Paste (clipboard → Ctrl+V), Type (xdotool keystrokes), or Copy-only.

All settings are in the Settings tab:

Config is stored at ~/.config/voicetotex/config.json.

┌─────────────────────────────────────────────┐
│  Electron Main Process                      │
│  ├── BrowserWindow (renderer)               │
│  ├── System Tray                            │
│  └── Floating Overlay                       │
└──────────────┬──────────────────────────────┘
               │ spawns
┌──────────────▼──────────────────────────────┐
│  Python Backend (WebSocket server)          │
│  ├── AudioCapture (sounddevice)             │
│  ├── Transcriber (faster-whisper)           │
│  ├── AudioDucker (pactl)                    │
│  ├── TextInjector (xdotool/xclip/xsel)     │
│  ├── HotkeyListener (pynput)               │
│  └── TranscriptionHistory (JSON)            │
└─────────────────────────────────────────────┘

voicetotex/
├── electron/          # Main process, preload, tray, overlay
├── src/
│   ├── css/           # Styles (dark/light themes)
│   ├── js/            # App logic, waveform, sounds, websocket
│   └── index.html     # Main renderer
├── backend/           # Python server, transcriber, audio
├── scripts/           # setup.sh, start.sh
├── build/             # Packaging scripts
├── docs/screenshots/  # App screenshots
└── package.json

• OS: Linux (Ubuntu 22.04+ recommended)
• GPU: NVIDIA with CUDA support (recommended for real-time transcription)
• CPU: Works on CPU too (set device to cpu in settings, slower)
• Python: 3.11+
• Node.js: 18+

MIT

• OpenAI Whisper
• faster-whisper / CTranslate2
• Electron