GitHub: noahpro99/3Dera Devpost: 3Dera on Devpost Demo: YouTube walkthrough
A hackathon project focused on making history learning immersive: enter AI-generated scenes in VR while listening to narrated context about the event. The AI 3D rendering happens on-the-fly based on the selected historical topic.
[vr, ai, education, 3d]
Why it stands out
- Won overall at HooHacks 2024 (as noted in the project README).
- Generates scenes from prompts instead of relying on pre-made fixed levels.
- 3D scenes are generated live from historical context rather than pre-rendered manually.
- Combines 3D scene generation with contextual narration for learning.
How it works
- Event understanding: uses GPT-4 to interpret a historical event and decide scene context.
- Scene construction: pulls objects from Objaverse and places/scales them in Blender with scripted coordinates.
- Environment generation: uses Hugging Face APIs for skybox and ground texture generation.
- Learning layer: uses Wikipedia + GPT-4 summarization to generate audio narration for the scene.
Technical Stack
- Frontend: React/TypeScript
- Backend: Python + FastAPI
- 3D pipeline: Blender scripting + Objaverse assets
- AI services: OpenAI, Hugging Face, Wikipedia API