Voice to text, reimagined.
A fast, native desktop app for speech recognition with AI-powered text processing.
Features • Installation • Usage • Building • License
Dictating text on a computer is often cumbersome:
- Cloud services are slow, expensive, and raise privacy concerns
- Built-in speech recognition is unreliable and hard to activate
- Professional software costs hundreds and is bloated with features
- No AI integration — transcribed text still needs manual editing
Vox solves these problems with an elegant approach:
- Global hotkey — Press
Option+M(Mac) orAlt+M(Windows) from anywhere - Local transcription — Whisper.cpp runs completely offline on your machine
- AI modes — Automatic grammar correction, translation, or custom prompts
- Instant clipboard — Result is immediately ready to paste
| Feature | Description |
|---|---|
| Global Hotkey | Start/stop recording from any app |
| Offline Transcription | Whisper.cpp — no cloud, no per-minute costs |
| Multi-Language | Auto-detect or manually select (EN, DE, FR, ES, ...) |
| AI Modes | Grammar fix, translation, or custom prompts |
| Statistics | Track your recordings, words, and characters |
| Background App | Runs quietly without cluttering your Dock |
Via OpenRouter:
- GPT-4o / GPT-4o Mini
- Claude 3.5 Sonnet
- Gemini Flash 1.5
- Llama 3.1 70B
Download the latest release for your platform:
| Platform | Download |
|---|---|
| macOS (Apple Silicon) | Download DMG |
| macOS (Intel) | Coming soon |
| Windows | Coming soon |
- Open the
.dmgfile - Drag Vox to your Applications folder
- Important: Open Terminal and run:
xattr -cr /Applications/Vox.app
- Double-click to open Vox
- Grant microphone permission when prompted
Note: The
xattrcommand is required because the app is not signed with an Apple Developer certificate. This removes the macOS quarantine flag that would otherwise block the app.
- Press Option+M (Mac) or Alt+M (Windows) to start recording
- Speak your text
- Press the hotkey again to transcribe
- Text is automatically copied to your clipboard
- Paste anywhere with Cmd+V / Ctrl+V
| Action | macOS | Windows |
|---|---|---|
| Start/Stop Recording | Option + M |
Alt + M |
| Cancel Recording | ESC |
ESC |
The hotkey can be customized in Settings.
- Normal — Pure transcription, no AI processing
- Fix Grammar — Corrects spelling and grammar mistakes
- Translate to English — Translates any language to English
- Custom — Create your own prompts for any text transformation
Note: AI modes require an OpenRouter API key. Normal mode works completely offline.
- Node.js 18+ and pnpm
- Rust (for Tauri)
- CMake and C++ compiler (for Whisper.cpp)
xcode-select --install- Visual Studio Build Tools with C++ workload
- Git
# Clone the repository
git clone https://github.com/hinrichb/vox.git
cd vox
# Install dependencies
pnpm install
# Setup Whisper (compiles whisper.cpp and downloads the model)
pnpm setup # macOS/Linux
pnpm setup:win # Windows (PowerShell)
# Development
pnpm tauri dev
# Production build
pnpm tauri buildBuild output: src-tauri/target/release/bundle/
┌─────────────────────────────────────────────────────────────┐
│ React/Next.js Frontend │
│ (Settings, Overlay UI, Audio Capture) │
└──────────────────────────┬──────────────────────────────────┘
│ IPC/Events
┌──────────────────────────▼──────────────────────────────────┐
│ Tauri Runtime (Rust) │
│ • Global Hotkey Handler │
│ • Window Management │
│ • Clipboard Integration │
└──────────────────────────┬──────────────────────────────────┘
│
┌────────────────┼────────────────┐
▼ ▼ ▼
┌──────────┐ ┌──────────┐ ┌──────────┐
│ Whisper │ │OpenRouter│ │ System │
│ .cpp │ │ API │ │ (Mic, │
│ (local) │ │(optional)│ │Clipboard)│
└──────────┘ └──────────┘ └──────────┘
- Frontend: React 19, Next.js 16, Tailwind CSS
- Backend: Tauri 2, Rust
- Transcription: Whisper.cpp (ggml-base model)
- AI: OpenRouter API
MIT
Built with Tauri, React, Next.js, and Whisper.cpp