YouTube Video Summarizer

A web application that generates intelligent summaries of YouTube videos using AI.

The application uses React for the frontend, FastAPI for the backend, MLX Whisper for audio transcription, and MLX-LM with the Llama-3.2-3B-Instruct-4bit model for summary generation.

Prerequisites

• Python 3.11+
• Node.js 22+
• macOS with MLX support (for AI)

Installation and Usage:

1. Clone the repository
2. Start the server

   cd backend
   python -m venv venv
   source venv/bin/activate
   pip install -r requirements.txt
   uvicorn main:app --host 0.0.0.0 --port 8000 --reload
   

3. Start the client

   cd frontend
   npm install
   npm run dev
   

4. Open the application in your browser

   http://localhost:5173
   

5. Paste the YouTube video URL in the input field
6. Click on the "Summarize" button
7. Wait for the summary to be generated (the process may take a few minutes)
8. Read the generated summary

Features

• ✅ Automatic audio extraction from YouTube videos
• ✅ Audio transcription to text with MLX Whisper
• ✅ Intelligent summary generation with MLX-LM
• ✅ Modern and responsive user interface
• ✅ Real-time progress tracking
• ✅ Multilingual support (automatic language detection)
• ✅ Asynchronous processing for better performance

Architecture

• Frontend: React + TypeScript + Vite + shadcn/ui
• Backend: FastAPI + Python
• AI: MLX Whisper (transcription) + MLX-LM (summaries)
• Video extraction: yt-dlp