Laxcorp-Research/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.

What it solves

Project Raven is an open-source meeting copilot designed to capture, transcribe, and provide AI-driven assistance during and after virtual meetings. It solves the problem of manual note-taking and the lack of real-time guidance during calls by providing a stealthy overlay for AI suggestions and automated post-call summaries.

How it works

Raven is an Electron desktop application that captures both microphone and system audio using native helpers (ScreenCaptureKit/CoreAudio on macOS and WASAPI on Windows). It employs GStreamer for local echo cancellation to ensure clean transcription streams for both the user and the other participants.

Transcription is handled via Bring-Your-Own-Key (BYOK) integration with Deepgram or AssemblyAI. AI assistance is powered by Anthropic or OpenAI, which processes the live transcript, optional screenshots, and locally stored documents. Data is kept private: recordings and notes are stored in a local SQLite database, and RAG (Retrieval-Augmented Generation) is performed on-device using the MiniLM embedding model.

Who it’s for

It is built for professionals who attend frequent virtual meetings on platforms like Zoom, Google Meet, Microsoft Teams, and Webex and want AI-powered insights and summaries without sending their data to a third-party hosted backend.

Highlights

  • Dual-stream capture: Separates "You" and "Them" for higher transcription accuracy.
  • Stealth overlay: An AI assistant interface that is invisible to screen-sharing software.
  • Sovereign data: Local SQLite storage and BYOK AI/STT providers; no Raven servers are used.
  • Local RAG: Ability to upload documents (.txt, .md, .pdf, .docx) for context-aware AI assistance during calls.
  • Automated post-call workflow: Generates structured action items, follow-up email drafts, and talk-ratio analysis.
  • Meeting auto-start: Detects meeting windows to prompt or automatically begin recording.

Related

  • Project
    jasoncheng7115/jt-live-whisper

    A fully local AI toolkit for real-time speech transcription, translation, and meeting summarization that works across any system audio source without cloud APIs.

  • Project
    Blueturboguy07/cue

    An open-source AI copilot that floats over your screen and remains invisible during screen shares, providing real-time assistance for meetings and coding.

  • Project
    EverMind-AI/Raven

    Raven is an open-source, self-improving agent harness that integrates long-term memory, proactivity, and local tracing for long-running AI agent workflows.

  • Project
    yazinsai/OpenOats

    A real-time meeting note-taker for macOS that transcribes calls locally and surfaces relevant talking points from your personal notes using RAG.

  • Project
    fastrepl/anarlog

    A privacy-first, local-first AI meeting notepad that records audio directly from your device to provide transcriptions and summaries without needing a bot to join the call.