CelebLearn started as a sleep-deprived hackathon idea at uOttaHack — what if your favorite celebrity could teach you calculus?
the problem
At the time, I was in high school watching everyone around me completely checked out from classes because of senioritis. The material itself wasn't necessarily the issue, people just weren't engaged. Around the same time, deepfake technology was blowing up online, mostly being used for memes and misinformation. Our team thought it would be interesting to use that same technology for something educational instead.
the idea
The idea was simple but ridiculous, upload a textbook page, choose a celebrity, and get an AI-generated lip-synced video of them teaching the concept back to you. Imagine Drake explaining derivatives or Morgan Freeman teaching thermodynamics. The content stays educational, but the delivery becomes entertaining enough to actually hold attention.
what i built
I worked on the FastAPI backend that coordinated the entire pipeline. We had to connect OCR, script generation, voice synthesis, lip syncing, video processing, transcription, and AI feedback systems into one cohesive workflow, all within 24 hours.
tech stack
The project used OCR to extract textbook content, OpenAI models to generate explanations and assess student understanding, Whisper for speech transcription, and the Sync Labs Lip Sync API to create realistic deepfake-style educational videos. After watching the lesson, users could record themselves explaining the concept back, and the AI would analyze their response, identify gaps in understanding, and provide feedback.
what i learned
One of the biggest things I learned from this project was how difficult it is to orchestrate multiple AI systems together in real time. Every stage depended on the previous one succeeding: PDF upload, OCR extraction, script generation, audio processing, lip syncing, rendering, and assessment. If one API failed, the whole pipeline broke. Debugging chained asynchronous systems at 3am on no sleep was genuinely chaotic, but also one of the fastest ways I've ever learned.
Hackathons also forced me to learn quickly and prioritize aggressively. A few of us had never even built a React app before, so we were learning TypeScript, frontend state management, async API handling, and deployment while actively building the product. There wasn't time to overthink things. We just had to figure problems out as we went.
More than anything, CelebLearn taught me that the best projects are usually the ones that sound slightly ridiculous at first. We started with "what if Drake taught calculus?" as a joke, and somehow ended up building a full AI-powered educational platform in a single weekend.
It was chaotic, sleep deprived, technically messy, and honestly one of the most fun things I've ever built.