MeetingCopilot is a real-time meeting copilot for Windows and macOS, providing live transcription, bilingual support, and first-person teleprompter answers using local or cloud ASR backends and BYOK LLMs.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (2 so far).
What it is
MeetingCopilot is a real-time meeting and interview copilot for Windows and macOS. It provides live transcription of the other side, first-person teleprompter answers grounded in the user’s resume, and capture protection. It supports multiple ASR backends (local FunASR streaming, local Whisper turbo, Alibaba Cloud fun-asr-realtime, MiMo per-segment) and can operate with OpenAI-compatible LLMs for auto-generated answers. It is designed to work with per-session resumes and offers per-meeting organization of transcript, questions, and materials.
How it works
The tool captures audio to transcribe the other side using selectable ASR backends. It streams ASR results and displays live subtitles while speech is ongoing. It can generate teleprompter-style answers by querying a BYOK LLM, with answers grounded in the imported resume and JD slots. It maintains per-session data (transcripts, questions, and materials) locally and can summarize interviews as a rolling memo. It supports bilingual language switching (zh/en) and can operate with UI language settings independent of answer language. It includes a per-platform setup for Windows and macOS, including platform-specific audio capture and stealth behavior.
Getting started
"git clone https://github.com/JWM0203/MeetingCopilot.git" "cd MeetingCopilot" "npm install" # postinstall applies patches/ (transformers.js patch — do not remove) "npm run build" # builds main + preload + renderer into out/ "npm start" # cross-platform; Windows can also use start.bat
First run:
- Open ⚙ Settings → pick the DeepSeek preset → paste your API key → save.
- Pick an ASR backend (default local streaming FunASR).
- Press ▶ Start to start transcription and auto-answers.
- Import resume/JD via 📄 / 📋 so answers are grounded.
Note: If npm / Electron downloads are slow in China, use a .npmrc with registry settings provided in the README.
Platform setup
Windows: docs/windows/SETUP.md; macOS: docs/macos/SETUP.md. Audio capture, stealth behavior, and guides are platform-specific.
ASR backends
- Local FunASR streaming (default)
- Local Whisper turbo
- Aliyun fun-asr-realtime
- MiMo per-segment
Local streaming FunASR (default)
One-time Python environment; app auto-spawns and reaps the sidecar at ws://127.0.0.1:10097. Model downloads on first run (~880 MB for paraformer, ~1.7 GB for Nano).
Privacy
API keys encrypted at rest; data stored per-user; no telemetry, no accounts, no server. Audio with local ASR backends never leaves machine; transcripts may stay local unless sent to BYOK LLM backend configured by user.
License
Apache License 2.0