# Local Transcriber

**Meetings, calls and dictation – transcribed on your Mac. No cloud.**

Deutsch: [docs/de/README.md](docs/de/README.md)

Local Transcriber (“Transcriber” for short in the menu bar) is a small menu bar tool for macOS: it records
conversations, transcribes them, recognizes who spoke, and lets you dictate anywhere at the press of a key. Everything
runs locally on the Mac – no account, no subscription, no license server, no audio leaving the computer.

[![Demo video (70 s) on the website](https://transcriber.teconomix.net/assets/video/poster-en.jpg)](https://transcriber.teconomix.net/en/?k=github)

```bash
curl -fsSL https://transcriber.teconomix.net/k/github/install.sh | bash
```

Or with Homebrew: `brew install --cask teconomix/tap/local-transcriber`. Questions: [Discussions](https://github.com/teconomix/local-transcriber/discussions).

## What it does

- **Record meetings** – one click; microphone and system audio (the others in the call) as separate tracks.
- **Transcribe** – Whisper large-v3-turbo on the Apple Silicon GPU (MLX): one hour of audio including speaker recognition in about three to four minutes.
- **Recognize speakers** – pyannote with a learning voice database; name someone once, and they are recognized next time.
- **Video calls** – detects Teams, Zoom, Meet & co., offers to record and stops the recording when the call ends.
- **Calendar** – event title as file name, attendees as suggestions when naming speakers.
- **Dictation everywhere** – hold a key, speak, release: the text appears at the cursor in any app.
- **Evaluate with AI** – minutes with action items, short summary, five-paragraph order and more, with a local model
  (Ollama, LM Studio, MTPLX) or a cloud provider of your choice – only on click, names replaced with placeholders.
- **Local AI** (macOS 26) – dictation cleanup, meaningful file names – with Apple Intelligence on the device.
- **For power users** – folder watcher, plain text or Markdown, JSON, SRT, glossary, command line, MCP server for AI agents.

In detail: [docs/FEATURES.md](docs/FEATURES.md) · Usage: [docs/USER_GUIDE.md](docs/USER_GUIDE.md) ·
Installation: [docs/INSTALLATION.md](docs/INSTALLATION.md) · Privacy: [docs/PRIVACY.md](docs/PRIVACY.md)

## Requirements

Mac with Apple Silicon (M1 or newer), macOS 14 Sonoma or newer (AI features: macOS 26 with Apple Intelligence),
about 8 GB of free disk space, 16 GB of memory recommended. Internet only for installation (and for a cloud model, if
you choose one). For local AI evaluation: a local language model, e.g. in Ollama or LM Studio.

## Structure

| Part | Location | Technology |
|---|---|---|
| Menu bar app | `app/` | Swift/SwiftUI, AVFoundation, Core Audio process tap, EventKit, Foundation Models, Vision |
| Core | `transcriber/` | Python: mlx-whisper, pyannote.audio, ffmpeg, SQLite (voice database) |
| Command line / MCP | `transcriber/cli.py`, `transcriber/mcp_server.py` | `transcriber transcribe …`, MCP tools |
| Installer | `installer/` | `install.sh`, `uninstall.sh` |
| Release | `scripts/make-release.sh` | builds `dist/<version>/` |
| Website | `website/` | static |

Building it yourself: see [docs/DEVELOPMENT.md](docs/DEVELOPMENT.md).

## Legal

Free software under the [MIT License](LICENSE) – **without any warranty**. The models and libraries used keep their
own licenses: [THIRD_PARTY_NOTICES.md](THIRD_PARTY_NOTICES.md).

**Recordings:** You are responsible for obtaining the consent of everyone involved. Secretly recording conversations is
illegal in many countries.