Inspiration
Educational videos are one of the biggest sources of learning today, but they often require viewers to spend a lot of time searching through long videos to find specific explanations. I wanted to make video learning more interactive and efficient.
The inspiration behind AI Video Tutor was to transform passive video watching into an AI-powered learning experience, where users can instantly understand concepts, ask questions, and interact with the video content like they would with a tutor.
What it does
AI Video Tutor turns any YouTube educational video into an AI-powered tutor. A user simply provides a YouTube video link, and the system: Extracts the video transcript Processes the content using AI Generates clear explanations and summaries Allows users to ask questions about the video Provides context-aware answers based on the video transcript Supports multiple languages, allowing users to learn content in their native language This multilingual capability helps students understand lectures even if the original video is in a different language. For example, a student can watch an English lecture but receive explanations in their native language, making learning more accessible.
How we built it
built AI Video Tutor using a combination of Google AI Studio, web technologies, and transcript extraction tools. Key components include: Google AI Studio to design and experiment with the AI workflow YouTube transcript extraction to obtain spoken content from videos AI models to summarize and answer questions based on the transcript A local backend server to process transcript data and handle API requests ngrok to expose my local server as a temporary public endpoint During development, many public servers were being blocked by YouTube when trying to access transcripts. To solve this, I exposed my local backend using ngrok, allowing the system to successfully fetch transcripts and process videos.
Challenges we ran into
One of the biggest challenges was retrieving YouTube video information through APIs when the requests were coming from public servers. Many public environments were blocked by YouTube, which caused API calls for video data and transcripts to fail.
To overcome this, I set up a local backend server and exposed it using ngrok, which allowed the system to fetch transcripts successfully without triggering the same restrictions.
Another challenge was capturing screenshots from the video. Screenshots are important because they can provide visual context to the AI model, helping it better understand what is being explained in the video. While the transcript issue has been resolved, automatic screenshot extraction is still a work in progress and will be implemented in the next iteration to improve the AI’s understanding of visual content.
These challenges highlighted the difficulties of working with external platforms, API restrictions, and multimodal AI inputs.
Accomplishments that we're proud of
I’m proud that I built a system that can turn any YouTube video into an interactive AI tutor. Key accomplishments include: Creating a working pipeline from video → transcript → AI explanation Successfully integrating AI Studio with a custom backend Solving YouTube API and server blocking issues Adding multilingual support, allowing students to understand lectures in their preferred language Building a simple interface where a single video link becomes a learning assistant
What we learned
Through this project, I learned a lot about: Integrating AI models into real applications Working with external APIs and their restrictions Using ngrok to expose local development environments Managing large text data pipelines for AI processing This project also helped me understand how AI can transform passive content like videos into interactive educational tools.
What's next for AI Video Tutor
I plan to extend AI Video Tutor with more powerful learning features such as:
- Creating animated artifacts, which help to explain video
- Automatic chapter and concept detection
- AI-generated study notes
- Interactive quizzes based on video content
- Timeline-based concept search
- Visual diagrams for complex topics
- My long-term vision is to build a platform where any video on the internet can instantly become an intelligent tutor.
Built With
- next.js
- react.js
Log in or sign up for Devpost to join the conversation.