Description
Getting your own room into a browser usually means a modeler spending days matching photos by hand, or a scanning app that hands back a lumpy mesh full of holes that nobody can edit. 3D Model Anything splits the job differently: the phone gathers the evidence and an AI coding agent does the modeling.
Scan the room with a LiDAR-equipped iPhone and the app exports the geometry along with the photos. Then ask a local-skills agent to run scan-to-threejs: it reads the measurements, writes an editable Three.js scene, and compares matched-camera renders with the original photographs, fixing things round by round. What you end up with is code you can keep changing, not a frozen mesh.
It comes from Wes Bos, and the repo ships his own office as a full example you can orbit in dollhouse, walking, overhead and wireframe views. He is upfront that it is an iterative reconstruction workflow rather than a one-click perfect model, and it currently focuses on rooms and interiors.
iPhone capture app: Scanning needs a LiDAR iPhone or iPad, while other devices can import and review scans. The source is in the repo and you build it with Xcode.
scan-to-threejs skill: Guides an agent through reading measurements, modeling, and comparing matched-camera renders with the original photos.
Three.js viewer: Run it locally with npm run dev, with dollhouse, walking, overhead, wireframe and knolling views.
Office demo: A reconstructed office ships with the repo and can be opened online.
Licensing: Code and skill are MIT licensed, and the demo images and textures have separate asset terms.
Scan the room with a LiDAR-equipped iPhone and the app exports the geometry along with the photos. Then ask a local-skills agent to run scan-to-threejs: it reads the measurements, writes an editable Three.js scene, and compares matched-camera renders with the original photographs, fixing things round by round. What you end up with is code you can keep changing, not a frozen mesh.
It comes from Wes Bos, and the repo ships his own office as a full example you can orbit in dollhouse, walking, overhead and wireframe views. He is upfront that it is an iterative reconstruction workflow rather than a one-click perfect model, and it currently focuses on rooms and interiors.
Features
iPhone capture app: Scanning needs a LiDAR iPhone or iPad, while other devices can import and review scans. The source is in the repo and you build it with Xcode.
scan-to-threejs skill: Guides an agent through reading measurements, modeling, and comparing matched-camera renders with the original photos.
Three.js viewer: Run it locally with npm run dev, with dollhouse, walking, overhead, wireframe and knolling views.
Office demo: A reconstructed office ships with the repo and can be opened online.
Licensing: Code and skill are MIT licensed, and the demo images and textures have separate asset terms.
