https://github.com/flexion/meeting-summarizer
https://github.com/flexion/meeting-summarizer
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/flexion/meeting-summarizer
- Owner: flexion
- Created: 2026-01-13T01:26:07.000Z (5 months ago)
- Default Branch: main
- Last Pushed: 2026-03-21T18:21:22.000Z (3 months ago)
- Last Synced: 2026-03-22T07:55:08.867Z (3 months ago)
- Language: Python
- Size: 354 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Agents: AGENTS.md
Awesome Lists containing this project
README
# Live Audio Transcription
Real-time audio transcription using faster-whisper and BlackHole for macOS. Capture and transcribe audio from your microphone, system audio, or any application. Includes a web UI and a headless Zoom bot for automatic meeting transcription.
> **Three ways to use it:** CLI for quick recordings, Web UI for a visual experience, or Zoom Bot for hands-free meeting capture.
## Prerequisites
- macOS (tested on macOS 10.15+)
- Python 3.10+
- [uv](https://docs.astral.sh/uv/): `brew install uv` or `curl -LsSf https://astral.sh/uv/install.sh | sh`
- System dependencies:
```bash
brew install ffmpeg portaudio blackhole-2ch
```
### BlackHole Setup
BlackHole routes audio between applications. To capture system audio (Zoom, YouTube, etc.):
1. Open **Audio MIDI Setup** (Applications > Utilities)
2. Click **+** > Create **Multi-Output Device**
3. Check **BlackHole 2ch** and your speakers/headphones
4. Set the Multi-Output Device as your system output
See [BlackHole docs](https://github.com/ExistentialAudio/BlackHole) for more details.
## Installation
```bash
git clone
cd meeting-summarizer
brew install ffmpeg portaudio blackhole-2ch # system deps (macOS)
make setup # creates .venv with uv
source .venv/bin/activate
make install # production deps
make install-dev # dev deps + pre-commit hooks (optional)
```
> **Note:** `make install` and `make install-dev` will check for required system
> dependencies (portaudio, ffmpeg) and fail with a helpful message if they are missing.
Optionally copy `.env.example` to `.env` to customize settings (model size, compute type, etc.).
## Architecture
## Usage
### CLI
```bash
make run # record + transcribe live audio
# or: uv run python transcribe_live.py
# Transcribe an existing audio file
uv run python transcribe_live.py --transcribe-audio path/to/audio.wav
# Summarize a transcript
uv run python transcribe_live.py --summarize transcripts/transcript_*.txt
# Chat with a transcript
uv run python transcribe_live.py --chat transcripts/transcript_*.txt
```
### Web UI
```bash
make run-web # starts FastAPI server at http://127.0.0.1:8000
```
### Zoom Bot
```bash
make install-playwright # one-time browser install
# Join a meeting (headed mode, runs until Ctrl+C)
make run-audio-test URL="https://zoom.us/j/123" HEADED=1
# With a time limit
make run-audio-test URL="https://zoom.us/j/123" HEADED=1 DURATION=60
# Join a specific breakout room
make run-breakout-test URL="https://zoom.us/j/123" ROOM="Room 1"
```
## Development
```bash
make format # ruff format + autofix
make lint # ruff check
make type-check # mypy
make test # pytest with coverage
make audit # pip-audit
make clean # remove caches and build artifacts
```
## Configuration
Settings are controlled via environment variables (`.env` file):
| Variable | Default | Description |
|----------|---------|-------------|
| WHISPER_MODEL | large-v3-turbo | tiny / base / small / medium / large-v3 / large-v3-turbo |
| DEVICE | cpu | cpu / cuda |
| COMPUTE_TYPE | int8 | int8 / float16 / float32 |
| OUTPUT_DIR | transcripts | Where recordings and transcripts are saved |
| AWS_REGION | us-east-1 | AWS region for Bedrock summarization |
| BEDROCK_MODEL_ID | (see .env.example) | Claude model for summarization/chat |
| AUTO_SUMMARIZE | true | Auto-generate summary after transcription (true/false) |
Zoom bot settings: `PLAYWRIGHT_HEADLESS`, `PLAYWRIGHT_TIMEOUT`, `PLAYWRIGHT_DEBUG` (see `.env.example`).
## License
(To be determined)