MemoMorph

MemoMorph helps university students transform scattered lecture and group recordings into clear, organized notes. It combines overlapping content, fills in gaps, and highlights.

Comment

About MemoMorph

Inspiration

MemoMorph was inspired by a common student experience: recording lectures and group discussions in pieces.

In practice, students rarely have a complete recording. Instead, they gather voice messages from classmates, partial recordings from their own phones, and clips captured at different times. When exam time arrives, the issue isn’t a lack of information; it’s that everything is scattered, overlapping, and inconsistent.

Most existing tools focus on transcription or summarization, but they assume the input is clean. MemoMorph starts from a different idea: academic audio is naturally messy, and tools should be built for that reality.

What I Learned

Building MemoMorph taught me that working with AI is less about generating answers and more about creating the right structure around uncertainty.

I learned:

  • How to design AI-assisted workflows that reveal uncertainty instead of hiding it
  • How to translate messy human behavior into clear interaction models
  • How important pacing, visual clarity, and trust are when creating tools for students
  • That a good demo is not about showing features, but about guiding understanding

Most importantly, I learned that clarity is a design choice, not just a result of using AI.

How I Built the Project

MemoMorph is built as a web-based application focused on rapid iteration and clear user experience:

  • Frontend: Vite + React
  • ASR: Whisper, served through a lightweight Express backend
  • AI Reconstruction: Gemini API, called directly from the frontend
  • Core Logic:
    • Fragment-level processing
    • Detection of overlaps, duplicates, and complementary content
    • Context-aware reconstruction for lectures and group discussions

Instead of providing a single “final answer,” MemoMorph maintains a transparent pipeline:

  • Original transcript
  • Reconstructed structure
  • Clear markers for inferred or conflicting content

This lets students understand how results are formed, not just consume them.

Challenges Faced

One of the biggest challenges was not over-automating.

It was tempting to hide uncertainty and produce something that appeared perfectly clean. However, doing so would break trust, especially in academic contexts.

Other challenges included:

  • Designing merge logic that feels intuitive rather than like “AI magic”
  • Maintaining a simple interface while managing complex underlying logic
  • Fitting a meaningful story into a strict 3-minute demo format

Balancing power, clarity, and honesty was the hardest — and most valuable — part of the project.

Closing Thoughts

MemoMorph is not just a transcription tool. It seeks to rethink how students interact with imperfect information.

By treating fragmentation as a key issue rather than an edge case, MemoMorph helps turn messy academic audio into something students can genuinely study with.

Share this project:

Updates