Voice dictation and meeting transcription
without data ever leaving your machine.
100% local models = 100% privacy
Download for MacmacOS 14.0+ · Apple Silicon (M1+) · Free & open source
★ Star on GitHub to support the project
Hold Control to talk, release to transcribe and paste into any text field. Works everywhere.
Record calls with notes, transcript, and AI-generated summaries — saved as local markdown.
Local LLM removes filler words, fixes self-corrections, and cleans up your speech automatically.
All models run on your Mac via Apple Silicon. Nothing is uploaded, tracked, or stored in the cloud.
"Apple could not verify" warning? Go to System Settings > Privacy & Security, scroll down, and click Open Anyway. You only need to do this once.
Every core feature runs 100% on your Mac — verified by AI code review.
| Feature | Status | What was checked |
|---|---|---|
| Speech-to-text | ✓ Local | WhisperKit/FluidAudio inference, no audio sent anywhere |
| Text cleanup | ✓ Local | Qwen LLM runs on-device via LLM.swift |
| Meeting transcription | ✓ Local | Chunked transcription, markdown files on disk |
| File storage | ✓ Local | Markdown to local filesystem, no cloud sync |
| Analytics & telemetry | ✓ None | No tracking SDKs of any kind |
| Speech Model | Size | Best for |
|---|---|---|
| Whisper tiny.en | ~75 MB | Fastest, English only |
| Whisper small.en (default) | ~466 MB | Best accuracy, English only |
| Parakeet v3 | ~1.4 GB | 25 languages |
| Qwen3-ASR 0.6B | ~900 MB | 50+ languages, macOS 15+ |
| Cleanup Model | Size | Speed |
|---|---|---|
| Qwen 3.5 0.8B (default) | ~535 MB | Very fast (~1-2s) |
| Qwen 3.5 2B | ~1.3 GB | Fast (~4-5s) |
| Qwen 3.5 4B | ~2.8 GB | Full quality (~5-7s) |