Video Summarization with ChatGPT

2023.03. ~ 2023.06.
숭실대학교 23-1학기 소프트웨어프로젝트에서 진행한 프로젝트입니다.
ChatGPT와 WhisperX 를 사용하여 AI 영상 요약 모델을 구현하고, 인터페이스를 통해 사용자에게 영상 요약 서비스를 제공했습니다.

SW1821

AI model 강준규 @KangJunGyu
AI model 이원호 @WonhoLee97
Front-end 성나영 @sna0e
Back-end 조수현 @chopha

---

This project summarizes Youtube videos efficiently by utilizing ChatGPT.

By applying STT model, we can extract a text file as the caption of a video. Then, ChatGPT summarizes the script. It compares original text with condensed text by applying cosine similarity measurement, and calculates each accuracy of each lines. In conclusion, We select parts of the text with higher accuracy to edit the video.

The website provides service for video summarization

1. User submits a Youtube video in the form of URL.
2. AI model utilizing ChatGPT and WhisperX makes a summarized video.
3. Website shows the result of the video.
4. User can download it.

stacks

Environment

Config

Development

AI model

Library

Communication

Requirements

for building running the appication you need:

• Node js 18.13.0
• Npm 9.4.0
• Python 3.11.2
• PyTorch 1.13.1
• Torchaudio 0.13.1
• Torchvision 0.14.1
• FFmpeg-Python 0.2.0
• CUDA 11.4
• WhisperX v2

Installtion

$ git clone https://github.com/SW1821/Video_Summarization_with_ChatGPT.git
$ cd Video_Summarization_with_ChatGPT

Backend

$ cd server
$ npm install
$ npm run develop

Frontend

$ cd client
$ npm install
$ npm run start

AI

$ cd model
$ pip install -e .
$ python model.py