Spatial Summarizer - Devpost Submission

Inspiration

Traditional documents are barriers for millions of students with ADHD, dyslexia, and autism. We wanted to transform how people learn by making educational content visual, interactive, and multi-sensory—turning flat PDFs into explorable 3D worlds.

What it does

Spatial Summarizer converts any document (PDF, PowerPoint, text) into an interactive 3D experience with AI narration. Upload your biology notes, and explore a 3D brain where you can click on regions to hear explanations. The output is a single HTML file—shareable, offline-capable, and accessible.

How we built it

  • Backend: FastAPI with Python for document processing (PyMuPDF, python-pptx)
  • AI: Google Gemini analyzes content and designs 3D scenes using LangChain
  • 3D Rendering: Three.js primitives (no external models needed)
  • Audio: ElevenLabs TTS generates narrations
  • Frontend: Premium glassmorphism UI with vanilla CSS
  • Three generation modes: Model matching (V1), anatomical generator (V2), and LLM-generated primitives (V3)

Challenges we ran into

  • Getting the LLM to design meaningful spatial relationships instead of random placements
  • Balancing visual appeal with accessibility requirements
  • Generating complete, self-contained HTML files with embedded interactions
  • Creating content-aware particle systems for anatomical visualizations

Accomplishments that we're proud of

  • Built three different generation engines (V1, V2, V3) in one hackathon
  • V3 mode creates truly educational 3D scenes with zero external dependencies
  • Achieved full keyboard navigation and ARIA compliance
  • Generated scenes work completely offline after download
  • Premium UI that rivals commercial products

What we learned

  • LLMs can be spatial designers, not just text generators
  • Accessibility isn't a feature—it's a foundation
  • Three.js primitives are more flexible than pre-made models
  • Multi-sensory learning significantly improves retention

What's next for Spatial Summarizer

  • VR/AR export for immersive learning
  • Real-time collaboration features
  • Support for more formats (DOCX, Markdown, videos)
  • Custom voice selection and multilingual support
  • Mobile apps for iOS/Android
  • Teacher dashboard for creating classroom experiences

Built With

Share this project:

Updates