Inspiration
I love traveling virtually with Google Street View, on the browser and in VR. With World Models, I wanted to bring these static images to life, so that users can experience the richness of the real world, virtually.
What it does
Given a Google Maps Street View link, it generates a full game world with animated human characters and location-appropriate music, then displays this in WebXR for the user to traverse.
How we built it
World Labs Marble for world generation (Gaussian splatting), Tripo 3D for asset generation, Gemini for panorama image segmentation and object detection, Googe Maps API for Map Tile retrieval. Basically, I built a multi-step pipeline that integrates all of these automatically, with Claude.
Challenges we ran into
Lighting, rendering, asset generation settings.
Accomplishments that we're proud of
Getting a passable version of Trevi Fountain rendered!
What we learned
How World Models work in practice, and different image processing techniques for panoramas (jk, I just used Gemini to do it all for me).
What's next for LA55 Team Street View
I'd like to get a real job.
Log in or sign up for Devpost to join the conversation.