Inspiration

The idea for SumMate came from the frustration of having to watch long YouTube videos just to find key points or relevant information or just to find it was a clickbait all along . Whether it's for studying, research, or work, time is precious, and i wanted to create a tool that could save people time by summarizing videos quickly and accurately. With Chromes ai we can locally summarize videos without internet calls there i saw an opportunity to leverage these technologies to build a solution that would make consuming youtube content more efficient and productive.

What it does

It takes the transcript of the YouTube page with JavaScript, then cleans it and sends it to the Gemini Nano API where it processes it locally and displays it in the sidebar. We have different options to parse the text, and users can save summaries in the popup to view them later. This makes it easy to quickly understand video content without watching the entire video.

How we built it

I used Google Chrome's Build-in AI API documentation and the sample GitHub code as references, and utilized YouTube's built-in transcription functionality to get text as input. The extension is built using JavaScript for the core functionality, HTML/CSS for the user interface, and integrates with Chrome's extension APIs for seamless browser integration.

Challenges we ran into

First, I attempted to use a video-to-text API to achieve the transcription, but it proved inefficient and the text quality wasn't meeting requirements. After further research, I discovered a way to directly access YouTube's transcript using JavaScript, which provided much better results. Additional challenges included:

  • Managing the Chrome extension lifecycle and permissions
  • Optimizing the local AI processing to ensure quick response times
  • Creating a user-friendly interface that wouldn't interfere with the YouTube experience

Accomplishments that we're proud of

  • Built my first Chrome extension from concept to working prototype
  • Successfully implemented local AI processing without requiring external API calls
  • Created a clean, intuitive user interface that integrates seamlessly with YouTube
  • Developed an efficient method to extract and process video transcripts
  • Achieved fast processing times while maintaining accuracy in summaries using custom prompts

What we learned

This was my first time building a Chrome extension, and it was an exciting learning experience. I learnt

  • Chrome extension architecture and development workflow
  • Working with Google's AI APIs and local machine learning models
  • JavaScript DOM manipulation for extracting web content
  • Managing and handeling of extension and browser permissions

What's next for SumMate

  • Implement customizable summary lengths and more styles
  • Create a summary history feature with search functionality
  • Integrate with other video platforms beyond YouTube
  • Add collaborative features for sharing and discussing summaries

Built With

Share this project:

Updates