AI Lecture Notes

Inspiration

As an international student, I found myself constantly translating between English lecture content and my own understanding, while also converting messy phone photos of slides into something I could actually study from. Existing OCR tools just extracted flat text — they didn't preserve structure, formulas, or context.

What it does

AI Lecture Notes converts lecture photos, PDFs, and slides into structured, bilingual notes instead of plain OCR output. It offers a bilingual notes workspace with source comparison, formula-preserving translation, confidence labels for AI-generated content, and an AI tutor with seven guided actions to help students actually engage with the material instead of just reading it passively.

How I built it

I used Gemini's structured-JSON generation to turn unstructured lecture images/PDFs into clean, organized note structures. The frontend is built with Next.js and TypeScript, with Supabase handling authentication and row-level security so users' private files stay private. Stripe powers the subscription/billing layer.

Challenges I ran into

Getting the AI to reliably preserve mathematical formulas through translation was the hardest part — naive translation would often break LaTeX formatting or lose formula meaning entirely. I also had to design the confidence-labeling system carefully so users know when to double-check AI-generated content versus when they can trust it.

What's next for AI Lecture Notes

I'm continuing to iterate on the AI tutor's guided actions and exploring how to extend the multimodal ingestion pipeline to handle more lecture formats.

Built With

• api
• gemini
• next.js
• supabase
• typescript