YouTube Video Summarizer

A Streamlit app that fetches captions from any YouTube video and produces a concise summary using Hugging Face's facebook/bart-large-cnn model.

Features

• Accepts full YouTube URLs and extracts the video ID automatically.
• Retrieves transcripts via youtube-transcript-api, handling disabled or missing captions gracefully.
• Summarizes long transcripts with tokenizer-aware chunking to avoid model limits.
• Optional GPU acceleration when PyTorch detects CUDA.
• Displays key timestamps and allows downloading the generated summary.

Quick Start

# Create and activate the conda env (already set up in this repo)
conda activate d:\youtube_summ\.conda

# Install dependencies (if needed)
python -m pip install -r requirements.txt

# Launch the app
streamlit run app.py

Then open the provided local URL in your browser (default: http://localhost:8501), paste a YouTube link, and hit Summarize.

Project Structure

• app.py – Streamlit UI orchestrating extraction, summarization, and display.
• extractor.py – Helpers to parse video IDs and fetch transcripts.
• summazier.py – Summarization logic built on Hugging Face transformers.
• requirements.txt – Python dependencies.
• README.md – You are here.

Notes

• Hugging Face caches models under your user profile; the first run downloads ~1 GB.
• Some videos block transcript access; the app surfaces clear error messages in those cases.
• To avoid committing local environments, keep .conda/ in .gitignore (already configured).

Happy summarizing! 🎥📝