# open-wispr

> Free, open-source voice dictation for macOS. Powered by Whisper AI, runs 100% on-device.

open-wispr is a lightweight macOS app that provides push-to-talk voice dictation. Hold a hotkey, speak, release — transcribed text appears at your cursor in any app. It uses OpenAI's Whisper model running locally via whisper.cpp, so no audio or text ever leaves the machine.

## Key Facts

- **Price**: Free and open source (MIT license)
- **Platform**: macOS (Apple Silicon and Intel)
- **Privacy**: 100% on-device. Zero network requests. Default maxRecordings=0: temp file + immediate delete, and startup clears any older retained recordings. Set maxRecordings 1–100 to enable Recent Recordings.
- **Install**: `brew install human37/open-wispr/open-wispr` or guided installer via `curl -fsSL https://open-wispr.pages.dev/install.sh | bash`
- **Source**: https://github.com/human37/open-wispr
- **Website**: https://open-wispr.pages.dev/

## How It Works

1. Hold the Globe key (or a custom hotkey)
2. Speak naturally
3. Release — text is typed at your cursor via macOS Accessibility

## Features

- Works in any app (text editors, browsers, Slack, Terminal, etc.)
- Multiple Whisper model sizes (tiny, base, small, medium, large)
- Configurable hotkey, language, and model via `~/.config/open-wispr/config.json`
- Runs as a background Homebrew service
- No GUI, no menubar icon, no overhead

## Comparison to Alternatives

| Feature | open-wispr | Wispr Flow | Superwhisper | MacWhisper |
|---|---|---|---|---|
| Price | Free | $15/mo | $8.49/mo | $69 |
| Open Source | Yes | No | No | No |
| 100% Local | Yes | Yes | Yes | Yes |
| Works in Any App | Yes | Yes | Yes | No (copy/paste) |
| Install Method | Homebrew | DMG | DMG | App Store |

## Configuration

```json
{
  "model": "base",
  "language": "en",
  "hotkey": "globe",
  "max_duration": 30
}
```

Config file location: `~/.config/open-wispr/config.json`

## System Requirements

- macOS 13 (Ventura) or later
- Apple Silicon (M1+) or Intel Mac
- ~150 MB disk space for the default Whisper model
- Microphone and Accessibility permissions