Inspiration
During school, we often watch long YouTube lectures or online classes, but it’s hard to take detailed notes while staying focused. Our team wanted a tool that would automatically create concise notes and flashcards from any educational video, making studying faster and smarter. We were inspired by the way generative AI can summarize content, but we wanted to combine it with a custom workflow that focuses specifically on students’ learning needs.
What It Does
Our application allows a user to paste a YouTube link or upload a recorded lecture, and it will: Extract the audio from the video. Transcribe the speech into accurate text Use a Generative AI model to produce: A clean, organized summary of the lecture. A set of flashcards in a Q&A format for quick review. This gives students instant study materials without having to manually pause and write notes.
How We Built It Audio Processing – We used yt-dlp to securely download audio when a YouTube link is provided, or allow direct file uploads.
Transcription –
- We used OpenAI and youtube transcript API keys to scrape the transcript of the copied link of the video
- The Gemini model we used has to go through the transcript and then follow the procedures after -After analyzing the transcript, the model generates: a) A summary of the video that helps the user understand the concept b) Some flashcard like questions about the video to assess the understanding. Also goes deeper.
Challenges We Ran Into IP Bans: Initially, we tried scraping YouTube transcripts directly using HTTP requests, which resulted in a temporary IP ban. We had to pivot to a safer and more reliable solution.
Accuracy vs. Speed: -Balancing transcription accuracy with processing time was tricky, especially for longer videos.
Prompt Engineering: -Getting the AI to output well-structured notes and usable flashcards required careful tuning of
prompts.
What We Learned
- How to integrate speech-to-text technology (Whisper) with Generative AI for real-time educational tools.
- The importance of using official APIs and ethical data access to avoid bans and scraping issues.
- How to design a workflow that students can easily use for productive studying.
What’s Next We plan to add:
Multilingual Support for non-English lectures.
Smart Highlighting to let users select key concepts to focus on.
Export Options to save flashcards directly into platforms like Anki or Quizlet.
Generative AI (Gemini / OpenAI GPT) – for summarization & flashcard creation
yt-dlp – for audio extraction from YouTube
HTML / CSS / JavaScript – for the web interface
Python (Flask/FastAPI) – backend processing
Built With
- css
- flask
- genai
- html
- javascript
- python
- ytdlp
Log in or sign up for Devpost to join the conversation.