Offline voice-to-text for macOS.
Speak, friend, and enter.
- 🔒 100% Offline — Your audio never leaves your machine. No cloud, no tracking, no analytics.
- 🎯 Two ASR Engines — NVIDIA Parakeet (state-of-the-art, 25 languages) and OpenAI Whisper. Both run locally.
- ⌨️ Global Hotkey — Press your shortcut in any app, speak, and text appears at your cursor.
- 🎙️ Wake Word — Say "Mimir, start recording" — hands-free dictation.
- 📁 File Transcription — Drop audio/video files (WAV, MP3, FLAC, OGG, MP4) and get text in seconds.
- 🤖 MCP Server — Connect Mimir to Claude Desktop, Cursor, or any MCP-compatible AI agent.
- ⚡ Built with Rust — Fast, lightweight, optimized for Apple Silicon.
Windows & Linux — coming soon.
┌─────────────────┐ ┌──────────────┐ ┌─────────────────┐
│ │ │ │ │ │
│ Global Hotkey ├────►│ │ │ Parakeet ASR │
│ │ │ ├────►│ (ONNX Runtime) │
│ Wake Word ├────►│ Mimir │ │ │
│ │ │ (Rust) │ ├─────────────────┤
│ File Drop ├────►│ │ │ │
│ │ │ ├────►│ Whisper │
│ MCP Server ├────►│ │ │ (Local) │
│ │ │ │ │ │
└─────────────────┘ └──────────────┘ └─────────────────┘
Input Tauri + CLI ASR Engines
Mimir provides a built-in MCP server for AI tool integration:
{
"mcpServers": {
"mimir": {
"command": "mimir",
"args": ["mcp"]
}
}
}Connect it to Claude Desktop, Cursor, or any MCP-compatible client. Transcribe audio right from your AI conversations.
Mimir is currently in early access. We're actively fixing bugs and polishing the experience.
- macOS (Apple Silicon)
- Parakeet & Whisper engines
- Global hotkey & wake word
- MCP server
- Auto-updater
- Windows support
- Linux support
- Open source release 🎉
Open source is coming. We're stabilizing the codebase and squashing bugs first. Once the core is solid, we'll open everything up. Stay tuned — star the repo to get notified!
Visit stensmir.github.io/mimir for more info, screenshots, and download links.
Found a bug? Have a feature request? Leave feedback — help us make Mimir better.
Built with ❤️ and Rust. Designed for privacy.

