Voice-to-text for developers featuring AI-powered cleanup and detailed analytics.
| Component | Directory | Stack |
|---|---|---|
| Python CLI | src/vibetotext/ |
Python 3.9+, Whisper.cpp, sounddevice, pynput |
| Windows native app | windows-native/ |
C# .NET 9, WPF, NAudio, Whisper.net |
| macOS native app | macos-native/ |
Swift 5.9, Metal, macOS 14+ |
| Electron history app | history-app/ |
Electron 28, better-sqlite3, D3.js |
All implementations share the same SQLite database at ~/.vibetotext/history.db.
Multi-Mode Hotkeys
Ctrl+Shift— Raw transcriptionCmd+Shift— Greppy mode with semantic code searchAlt+Shift— Cleanup mode (AI refines rambling into clear prompts)Cmd+Alt— Plan mode (generates structured implementation plans)
Fast Local Transcription
- Whisper.cpp for 2-4x faster transcription than Python Whisper
- Technical vocabulary bias for programming terms
- Auto-paste to cursor
Press Cmd+Comma (macOS) or Ctrl+Comma (Windows) to open the History & Settings window.
- Streaks & Personal Records — Track your current streak, best WPM, most words/day, and longest session.
- Topic Speed & Mood — See how fast and positive you are across topics like Testing, Planning, Documentation, and more. Bar colors shift from negative to positive sentiment.
- Daily Goal Progress — Set daily and weekly word targets and track completion.
- Activity Heatmap — GitHub-style hourly/yearly view of when you dictate most.
- Peak Hours & Words Over Time — Visualize your productivity patterns and dictation volume trends.
- Filler Words & Vocabulary Diversity — Monitor filler word usage and track your unique word count and richness score.
- Recent History — Review and copy previous transcriptions.
- Microphone Selection — Switch audio input devices directly from the UI.
A living 3D world that reacts to your voice in real time. As you dictate, a procedural planet grows with villagers, buildings, crops, and a tree whose leaves are your most-used words. A cosmic entity watches from a black hole in the sky — and if your sentiment turns negative, it attacks.
Hotkeys
Cmd+Ctrl+G— Open the Word Galaxy visualization
How it works
- Sentiment-driven behavior — Your words are analyzed in real time. Positive speech keeps the world peaceful; negative sentiment triggers the cosmic entity to charge and fire lasers at your village.
- Procedural planet — Villagers (farmers, scholars, builders, guards) and buildings populate a 3D planet that grows as you talk.
- Word tree — Your top 500 words are assigned to leaves on a procedural tree that grows during the intro sequence.
- Word nebula — Recent transcriptions float as text in a nebula cloud. Common words migrate from the nebula to the tree.
- Seasons & day/night — A 15-second day/night cycle with shifting sky colors, dynamic lighting, and fireflies at night.
- GLB export — Export generated 3D entities for use in external tools.
Requires macOS 14+ (Sonoma). The cosmic visualization is part of the native macOS app in
macos-native/, built with Swift and Metal. Not available on Windows or Linux.
pip install -e .
pip install -e ".[gemini,dev]" # with Gemini + dev dependenciesOptionally set GEMINI_API_KEY in a .env file to enable cleanup/plan modes. You can copy the .env.example file and then add your key.
cd windows-native
build.bat # or: dotnet build src/VibeToText/VibeToText.csprojcd macos-native
swift buildRequires macOS 14+ (Sonoma) and Swift 5.9+.
These scripts package the Python CLI into standalone executables via PyInstaller:
# Windows (from project root) → dist/vibetotext-engine.exe, dist/vibetotext-ui.exe
build_windows.bat
# macOS
bash packaging/macos/build_macos.sh
# Linux
bash packaging/linux/build_linux.shExecutables will be in the dist/ folder. See packaging/ for platform-specific configs.
vibetotext # Start with default hotkeys
vibetotext --model base # Use specific Whisper modelstart-all.shstop-all.sh