Abdullah-Rizwan's picture
Update README.md
639fb99 verified
|
Raw History Blame Contribute Delete
2.87 kB

A newer version of the Gradio SDK is available: 6.32.0

Upgrade
metadata
license: mit
title: Audio Lecture to Notes Generator
sdk: gradio
emoji: πŸ“š
colorFrom: yellow
colorTo: yellow
short_description: An agent that converts lecture audio into clean Notes

πŸŽ“ AI Lecture Notes Generator

Transform lecture audio into structured, comprehensive notes using AI.

Features

  • πŸŽ™οΈ Transcribes lecture audio using OpenAI's Whisper Large V3 model
  • πŸ“ Generates well-structured markdown notes using Meta-Llama-3.1-8B
  • 🧠 Identifies key concepts, points, and takeaways automatically
  • πŸš€ Simple web interface built with Gradio
  • ⚑ Optimized with 4-bit quantization for efficient inference

Demo

How It Works

  1. Audio Transcription: Converts lecture audio to text using Whisper Large V3
  2. Note Generation: Processes transcription through Llama 3.1 to create structured notes
  3. Format Cleaning: Ensures consistent markdown formatting with proper headers

Installation

# Clone the repository
git clone https://github.com/Abdulllah-Rizwan/AI_Lecture-s-Note-Generator.git
cd AI-Lecture-Notes-Generator

# Install dependencies
pip install requests torch bitsandbytes transformers sentencepiece accelerate gradio httpx==0.27.2

Usage

Option 1: Run the Jupyter Notebook

jupyter notebook "AI Lecture Notes Generator.ipynb"

Run all cells in the notebook to start the Gradio interface, then interact with the application through the displayed URL (typically http://127.0.0.1:7860/).

Option 2: Use the functions directly in your code

You can import the functions from the notebook into your own Python code:

# Import the necessary functions
from AI_Lecture_Notes_Generator import process_audio

# Generate notes from an audio file
notes = process_audio("path_to_lecture_audio.mp3")
print(notes)

Requirements

  • Python 3.8+
  • PyTorch
  • Transformers
  • Gradio
  • Hugging Face account with appropriate model access
  • CUDA-compatible GPU (recommended)

Output Format

The generated notes follow this structure:

# Topic: [Lecture Topic]
## Instructor
[Instructor name if detected]
## Summary
[Concise summary of the lecture content]
## Key Concepts
- [Important concept 1]
- [Important concept 2]
...
## Key Points
- [Major point 1]
- [Major point 2]
...
## Takeaways
- [Takeaway 1]
- [Takeaway 2]
...

Limitations

  • Best suited for academic lectures with clear structure
  • Audio quality significantly impacts transcription accuracy
  • May require fine-tuning for specialized terminology

License

MIT License.

Acknowledgements

  • OpenAI for the Whisper model
  • Meta for the Llama 3.1 model
  • Hugging Face for model hosting and libraries
  • Gradio for the web interface framework