Automated content digestion pipeline converting long-form video content into concise, actionable summaries.
In the age of information overload, consuming hour-long technical talks or tutorials is time-consuming. VideoSummarizer is a Python-based utility tool designed to scrape, process, and distill YouTube video transcripts into structured notes using Large Language Models (LLMs).
It solves the "Too Long; Didn't Watch" problem by automating the retrieval and summarization process.
- Transcript Extraction: Automatically fetches subtitles/transcripts from YouTube videos via API.
- Smart Chunking: Handles token limits by intelligently splitting long transcripts into manageable context chunks for the LLM.
- Semantic Summarization: Uses Generative AI to extract key points, ignoring filler content.
- Markdown Export: Outputs clean, readable notes ready for Notion or Obsidian.
- Language: Python 3.x
- Core Logic:
youtube-transcript-apifor data retrieval. - AI Integration: OpenAI API / Local LLM (configurable).
- Data Processing: Custom token management for optimal context window usage.
- Clone the repo:
git clone [https://github.com/icantdo/VideoSummarizer.git](https://github.com/icantdo/VideoSummarizer.git)
- Install dependencies:
pip install -r requirements.txt
- Run the summarizer:
python main.py --url "[https://youtube.com/watch?v=example](https://youtube.com/watch?v=example)"
- Adding support for local Whisper models for videos without transcripts.
- Integrating a simple Streamlit UI for non-technical users.
- Vector Database integration to "chat" with the video content.