NoteFlow AI

• 

  Start up page

• 

  Before Transcription

• 

  During Transcription

• 

  After Transcript and Note Generation part 1

• 

  After Transcript and Note Generation part 2

• 

  After Transcript and Note Generation part 3

Inspiration

As students, we know the struggle of keeping up with lectures while taking meaningful notes. Often, we miss important details or waste hours rewriting our notes. We wanted to build something that makes learning easier, faster, and smarter.

What it does

AI Lecture Note Taker lets you upload a lecture recording, automatically transcribes the audio using Whisper, and then summarizes it into structured notes with GPT-5 nano. The notes are neatly divided into sections like Main Points, Key Definitions, Examples, and Study Tips — making them immediately useful for studying.

How we built it

• Frontend/UI: Built with Streamlit, styled with custom CSS for a sleek and modern look.
  
• Transcription: Handled by OpenAI’s Whisper model, with audio preprocessing and chunking to handle longer files.
  
• Note generation: Uses OpenAI’s GPT-5 nano to summarize transcripts into structured, easy-to-read notes.
  
• Tech stack: Python, Streamlit, Whisper, OpenAI API, FFmpeg, Pydub.
  

Challenges we ran into

• Whisper transcription was slow at first — we optimized it with audio chunking.
  
• Making the UI visually appealing took extra CSS customization.
  
• Managing API costs while testing multiple transcripts was tricky.
  

Accomplishments we’re proud of

• Built a fully working end-to-end pipeline in under a hackathon timeframe.
  
• Created a clean, modern, and hackathon-ready UI with custom themes.
  
• Notes generated are actually useful for studying, not just summaries.
  

What we learned

• How to combine multiple AI models (Whisper + GPT) into one workflow.
  
• How to optimize transcription pipelines for speed.
  
• How to design with the user in mind — a simple interface makes AI much more accessible.
  

What’s next for NextFlow AI

• Multi-language transcription support.
  
• Export notes directly as PDF/Markdown.
  
• Adding collaboration features so students can share notes in real-time.
  

Built With

• ffmpeg
• open-ai-api-(gpt-5-nano)
• pydub
• python
• streamlit
• subprocess
• tempfile
• whisper