MeetScript

A tool set to transcribe and translate meetings in real-time.

Important Notice: Zoom has released official transcription and translation functionality, so this repository's development is postponed although it is fully functional.

Overall Structure

The application uses:

• FastAPI and Uvicorn for backend
• SQLite and Peewee for database
• OpenAI Whisper for transcription
  • And henceforth PyTorch for neural computation
• React and ChakraUI for frontend

In order to run the server,

• Python (minimum v3.9)
• Node.js (minimum v16) needs to be installed. Use
• cd ms-server; pip install -r requirements.txt
• cd ms-frontend; npm i to install all the requirements.

If you choose to use Whisper api and you have available GPU, installing CUDA and CUDA-version PyTorch will be a good option.

Deployment

Launching Directly

After the requirements are set, use

• ./launch.sh to launch the server with static files being mounted.
• ./launch-only.sh to launch the server without static files.

Using Reverse Proxy

If Nginx is installed, use ./deploy.sh to configure the site and launch the server.

Development

TODO