CatchUp

Inspiration

Meetings and classes move fast, and it’s easy to miss important details like decisions, deadlines, or key ideas. Students with hearing difficulties, attention challenges, or other disabilities often face extra barriers. We wanted a smarter way to capture and revisit information without losing focus, and CatchUp helps with this by transcribing speech to text.

What it does

Users can start by uploading a recording, so the app continuously listens in real time, ensuring every spoken detail is captured without requiring manual note-taking. The user can then ask AI questions about what was spoken during the recording, and read a recap at the end.

How we built it

We built a responsive interface with React, Next.js, and Tailwind CSS for live meetings, real-time transcripts, and AI Q&A. The frontend sends audio and queries to a FastAPI backend running on Uvicorn. Pydantic validates all data, and the Gemini API generates transcripts and analyzes context. Transcripts are stored in ChromaDB, enabling semantic search so users can ask questions and get precise, context-aware answers.

Challenges we ran into

-It was difficult to code it to transcribe the recording at first. -Resolving these issues required careful setup of credentials and improving communication between the frontend, backend, and Zoom services.

Accomplishments that we're proud of

Our app makes meetings searchable and easy to revisit by providing real-time transcripts. Built within 24 hours, it combines a full-stack frontend, backend, and AI services to capture live audio and convert it into accurate transcripts instantly. Users can also ask questions about meetings and receive precise, context-aware answers.

What we learned

We gained hands-on experience with live speech-to-text pipelines and AI transcription. Building the app taught us full-stack integration with React/Next.js and FastAPI, smooth data flow for live streams, and working with ChromaDB for semantic search. We also navigated challenges integrating the recording and learned to adapt quickly to new technologies while troubleshooting in real time.