Your invisible AI meeting assistant — Live captions and smart summaries, right in your browser.

Quick Start • How It Works • Features • API

ghostmeet silently captures audio from any browser tab — Google Meet, Zoom, Teams, or anything with sound — and transcribes it in real-time using Whisper. When the meeting ends, click Summarize and AI extracts key decisions, action items, and next steps.

It runs as a Chrome Extension side panel. Other participants can't see it. Like a ghost in your meeting. 👻

• 100% local — audio never leaves your machine
• No accounts — no sign-up, no cloud, no subscriptions
• Works everywhere — any tab that plays audio

• 🎙️ Real-time transcription — Whisper STT, updates every 10 seconds
• 📋 AI-powered summaries — Key decisions, action items, next steps
• 📎 Meeting context — Add agenda or attach files before the meeting
• 🔒 Self-hosted — your audio stays on your machine
• 🐳 One-command setup — docker compose up and you're ready
• 👻 Invisible — side panel UI, no one in the meeting knows

Browser Tab (Zoom / Meet / Teams)
    │
    ├── Chrome Extension captures tab audio (chrome.tabCapture)
    │
    ▼
WebSocket ──→ Local Backend (FastAPI + Whisper)
                  │
                  ├── Real-time STT ──→ Live Captions (Side Panel)
                  │
                  └── Claude AI (on demand) ──→ Meeting Summary

Everything runs on your machine. The only external call is to Claude API when you click Summarize (optional — transcription works without it).

• Docker (recommended) or Python 3.10+
• Chrome browser

git clone https://github.com/Higangssh/ghostmeet.git
cd ghostmeet

# Copy and edit config (add your Anthropic API key for summaries)
cp .env.example .env

# Start with Docker
docker compose up -d

Backend is ready when you see http://0.0.0.0:8877 in the logs.

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python -m backend

1. Open chrome://extensions in Chrome
2. Enable Developer mode (toggle in top-right)
3. Click Load unpacked
4. Select the extension/ folder from this repo
5. Pin the 👻 icon in your toolbar

1. Join a meeting — Open Google Meet, Zoom, Teams (or any tab with audio)
2. Click 👻 — Side panel opens on the right
3. Click ▶ Start — Live captions appear as people speak
4. Click ■ Stop — When the meeting ends
5. Click 📋 Summarize — AI generates a structured summary

That's it. No sign-up, no config, no cloud.

Set these in .env or docker-compose.yml:

Model size guide:

• tiny — fastest, least accurate (~75MB)
• base — good balance (recommended, ~150MB)
• small — better accuracy, slower (~500MB)
• medium / large — best accuracy, needs GPU

ghostmeet/
├── extension/              # Chrome MV3 Extension
│   ├── manifest.json       # permissions + side panel config
│   ├── background.js       # tab audio capture → WebSocket
│   ├── sidepanel.html/js   # live captions UI
│   ├── popup.html/js       # start/stop controls
│   └── icons/              # extension icons
├── backend/                # Python backend (FastAPI)
│   ├── app.py              # HTTP + WebSocket server
│   ├── audio_processor.py  # Whisper transcription pipeline
│   ├── transcriber.py      # faster-whisper wrapper
│   ├── summarizer.py       # Claude API integration
│   └── models.py           # session + segment models
├── assets/                 # logo, demo GIF
├── docker-compose.yml      # one-command deployment
├── Dockerfile              # backend container
└── requirements.txt        # Python dependencies

ghostmeet works as an OpenClaw skill. Control your meetings from chat.

# Install the skill
clawhub install ghostmeet

Then just ask your AI assistant:

• "Summarize my last meeting" → generates AI summary from latest session
• "How many meetings did I have today?" → lists all sessions
• "What was discussed?" → fetches full transcript
• "Extract action items" → pulls tasks from the summary

The skill handles session listing, transcript retrieval, and summary generation via the ghostmeet API. Recording start/stop is done through the Chrome Extension.

• Real-time transcription (Whisper)
• Chrome Extension side panel UI
• AI meeting summaries (Claude)
• Meeting context input + file attach
• Speaker diarization (who said what)
• Export to Markdown / PDF
• Multi-language support
• Agent Mode — AI speaks in the meeting for you

MIT