swe-web-app: Tower of Babble

Repository containing our web-application for SWE project (as a part of the coursework of IIT Tirupati), adhering to United Nations Sustainablility Goals

The idea of the project is given in the Team6 ProjectConcept.docx.

Importance of the project:

Allowing corporations to host our web-applications in their servers, we provide a free, easy and privacy-secure solution for translating and transcribing their data.

Uniqueness

No other solution exist outside of research domains, and that too do not provide a simplistic way to translate and transcribe text/audio for users

Long story short: We are building a website that can transcribe and translate input audio/video files to text/srt files using AI based models of Whisper.cpp and Fairseq

The major idea behind this is that the connection between express and python and the actual deployment of the models, which were poorly documented and for fairseq, poorly supported.

Contributions

Arup Biswas : Documentation and Tech Research (Existing solutions and usable tech)
Arvind Srinivasan : Docker, Model deployments in Python, ExpressJS bridge
Dhriti Chintakunta : User Interface development
Keshav Kumar Manjhi : MongoDB setup, ExpressJS setup, Middleware integration
Abhinav Gupta : User Interface Design, Express JS setup

Doc links

Project Concept : link
Requirements : link
Use Cases : link
Weekly Status Report : link
Design : link

To run the website

Compiling docker to create the image

Install nvidia-container-runtime or its equivalent for your distribution to use CUDA acceleration

set DOCKER_BUILDKIT=1
sudo docker build -t transcribe_server:latest Backend/

Run the following commands in order

cd Backend
sudo docker-compose up # if using CUDA GPU accelerated inference -- recommended for performance
sudo docker-compose -f compose-cpu.yml up # if using CPU

(in new tab)

cd Frontend
npm i 
npm start

Link to the repo: https://github.com/Aeromaster213/swe-web-app