🎓 AI Lecture Notes Generator

Automatically converts any YouTube lecture into structured academic notes in Markdown and PDF — just provide a URL.

---

📁 Project Structure

project/
│
├── main.py                  # Entry point — runs the full pipeline
├── config.py                # Path & title configuration, initialized once
├── transcript.py            # YouTube transcript extraction (captions or Whisper)
├── ai_agent_summarizer.py   # LLM-based summarization via Groq
├── markdown_to_pdf.py       # Converts markdown output to PDF via pandoc
├── header.tex               # LaTeX styling for PDF export
├── .env                     # API keys (not committed)
│
└── Outputs/
    └── <Video Title>/
        ├── Transcript.txt
        ├── AI_Summary.md
        └── AI_Summary.pdf

---

⚙️ How It Works

The pipeline runs in 3 stages when you provide a YouTube URL:

YouTube URL
    │
    ▼
[1] transcript.py       →   Extracts transcript (captions or Whisper fallback)
    │
    ▼
[2] ai_agent_summarizer.py  →   Sends transcript to Groq LLaMA-3.3-70b
    │                            Generates structured academic markdown notes
    ▼
[3] markdown_to_pdf.py  →   Converts markdown to PDF via pandoc + xelatex

---

🚀 Usage

1. Set your YouTube URL in main.py:

YOUTUBE_URL = "https://youtu.be/your_video_id"

2. Run the pipeline:

python main.py

Output files are saved to Outputs/<Video Title>/.

---

🔧 Installation

Prerequisites

• Python 3.9+
• pandoc
• xelatex (via TeX Live or MiKTeX)
• yt-dlp
• ffmpeg (required by Whisper)

Install Python dependencies

pip install -r requirements.txt

Set up .env

Create a .env file in the project root:

GROQ_API_KEY=your_groq_api_key_here

Get a free Groq API key at console.groq.com.

---

📝 AI Summary Format

The generated notes follow an academic structure:

• Numbered headings and subheadings
• Bullet-point explanations for each topic
• Expanded technical details for briefly mentioned concepts
• Properly formatted code snippets
• Comparison tables where applicable
• Full markdown formatting, PDF-ready

---

🔁 Transcript Extraction Logic

transcript.py tries two methods automatically:

Method	When Used
YouTubeTranscriptAPI	Video has captions available
OpenAI Whisper (base model)	No captions found — downloads audio and transcribes locally

---

📄 PDF Styling

PDF output is styled via header.tex using:

• 1-inch margins
• 1.2x line spacing
• Monospaced code blocks with a light grey background (DejaVu Sans Mono)

To customize, edit header.tex directly.

---

⚠️ Notes

• Re-running the pipeline on the same URL will overwrite all previous output files automatically.
• config.py must be initialized via config.init(url) before any other module is imported — this is handled by main.py.
• markdown_to_pdf.py uses header.tex resolved relative to the script's location, so it works regardless of where you run main.py from.