project-raven
Open-source AI meeting copilot - real-time transcription, echo cancellation, and AI assistance. Captures system audio + mic, cancels echo via WebRTC AEC3, trans
What is project-raven?
Open-source AI meeting copilot - real-time transcription, echo cancellation, and AI assistance. Captures system audio + mic, cancels echo via WebRTC AEC3, transcribes with Deepgram, and gives you Claude/OpenAI help during meetings. Runs locally on macOS and Windows.
Overview of project-raven
What it is
Project Raven is an open-source AI meeting copilot that runs locally on macOS and Windows. It captures system audio and microphone input, cancels echo using WebRTC AEC3, transcribes speech in real time via Deepgram or AssemblyAI, and provides AI assistance through Anthropic or OpenAI. All recordings and notes are stored in a local SQLite database; no data is sent to a Raven server. Users bring their own API keys for transcription and AI services.
Key features
- Dual-stream capture: system audio and microphone (macOS: ScreenCaptureKit + CoreAudio; Windows: WASAPI loopback + capture)
- Echo cancellation: GStreamer WebRTC AEC3 plus a residual echo gate
- Real-time transcription on two parallel streams (You and Them) using Deepgram nova-3 and/or AssemblyAI u3-rt-pro
- AI assistance from Anthropic or OpenAI via an overlay (Assist, recap, follow-up, custom prompts)
- Stealth overlay that hides the overlay and dashboard from typical screen-share APIs (Zoom, Meet, Teams, Discord)
- Local RAG per mode: upload .txt, .md, .pdf, or .docx; on-device embeddings with Xenova/all-MiniLM-L6-v2
- Session management: local SQLite storage, resume capability, incognito mode, post-call recap with action items, talk ratio, and export to Markdown or PDF
Use cases
- Real-time meeting assistance: get suggestions and answers during calls without leaving the meeting window
- Post-meeting summarization: automatically generate notes, action items, and follow-up email drafts
- Meeting history search: ask questions across all past meetings using local retrieval
- Privacy-conscious transcription: keep all data on your machine with your own API keys
Getting started
See the repository README for download and setup instructions.
License and pricing
This project is licensed under the MIT License and is open source.
Based on the project's public README and repository metadata.