Diorama Theater — gallery and WebXR scenes at the hackathon

Founder’s Inc Night Hack 2026

About

You give it 3 images—background, mid, foreground—and it generates a 3D scene.

I built a pipeline that turns flat images into layered XR dioramas.

What
It Does

Diorama Theater composites three flat PNG layers—background, midground, foreground—into a spatially separated theater scene with real parallax depth. Drop it into a browser for WebXR, send it to Vision Pro, or load the glTF in any 3D viewer. One CLI command; five export formats.

Built for
Vision Pro

and iOS Augmented Reality

Every scene exports a native RealityKit scene for visionOS. The same three PNG layers become spatially separated planes in an immersive diorama you can walk around in Apple Vision Pro — or hold in front of you in iOS AR.

How it works

01

Provide 3 layers

Pass three PNGs: background, midground, foreground. Any flat art works—illustrations, photos, cut-outs.

02

Pipeline composites

Each plane is placed at a fixed Z depth (BG −4m, MID −2.5m, FG −1.2m) inside a 3D coordinate space.

03

Five export formats

glTF for any 3D viewer, A-Frame WebXR for the browser, USD/USDZ for Apple tools, and a RealityKit Swift file for VisionOS.

04

Publish to the web

The gallery auto-generates from a scene manifest and deploys to GitHub Pages with one command.

Stack

Event

Built at the Build Night Hackathon hosted by Founder’s Inc by romyilano.

Thanks

Big thanks to the Vibe Coding Nights crew for the sessions that got me sharp on building with agents. This project wouldn’t exist without those nights.

Technical — Spatial layout

BG   Z=−4.0m   3.2×1.8m   Y=2.0m
MID  Z=−2.5m   2.4×1.35m  Y=2.0m
FG   Z=−1.2m   1.6×0.9m   Y=2.0m
Eye  Y=1.6m