Project Video to PPT

A project to experiment with LLM's generate simple presentation from videos.

---

Models

Models	Templates
OPENAI/gpt-3.5-turbo	SUMMARY, TITLE GENERATION, SUMMARY V2
MBZUAI/LaMini-Flan-T5-248M	SUMMARY, TITLE GENERATION, QA CHAT
google/flan-t5-large	SUMMARY, TITLE GENERATION
sshleifer/distilbart-cnn-12-6	SUMMARY
fabiochiu/t5-small-medium-title-generation	TITLE GENERATION

SUMMARY - Generates summary of the video TITLE GENERATION - Generates title for the video SUMMARY V2 - Generates summary of the video with more context. ( COSTLY/unoptimized ) QA CHAT - A chatbot interface using vecorstores to quickly search for the related data and fed to the LLM's to generate the answer for the question.

Checkout docstrings for more info

Installation

apt install git -y
git clone https://github.com/SaicharanKandukuri/project-video-to-ppt
cd project-video-to-ppt
pip install -r requirements.txt

Usage

For basic command line usage to generate PPT

python3 main.py -v <video_id> --no-chapers

GUI interfaces

All the task have a seperate GUI for this time being.

Video To PPT

python3 main.py --gui-web

Generates slides with video summary and convert to presentation or pdf

for any given video, the project will generate a summary of the video and will create a PPT/PPTX with different slides covering the summary of the video.

The video subtitles are fetched from youtube and fed to the LLM's to generate the summary of the video then the summary is used to create the PPT/PPTX with marp.

Video TO Question-Answer interface

python3 main.py -v VIDEO_ID --questions-mode

QA interface uses vectorstores to quickly search for the related data and fed to the LLM's to generate the answer for the question.

image source & refer about vector stores: lanchain.com

Diagram generation

Runs with GPT 3 for better reasoning capabilities

python3 main.py --diagram-gen