HouseTwin
Turn ordinary phone and drone photos of a property into an interactive, browser-viewable 3D model — no scanner, no LiDAR rig. The same reconstruction core behind ASI Labs spatial twins, shown as an honest early preview.
What it is
HouseTwin reconstructs a property in 3D from ordinary RGB photos or video using a classical photogrammetry pipeline (feature matching + structure-from-motion) trained into a 3D Gaussian splat, then serves it as an interactive precomputed scene that runs in a browser with no app install. A companion iOS app captures synchronized RGB + LiDAR-depth frames with per-frame pose.
The demonstrated result is a single exterior capture: 245 photo frames registered into one camera model (~1.43 px mean reprojection error), cleaned to ~829,000 Gaussians, rendered in a browser viewer that passed its smoke test.
Where it fits
Capturing and viewing a splat in a browser is now commoditized (Polycam, Scaniverse, the free SuperSplat viewer), so that layer isn't the value. The genuine whitespace is a unified exterior + interior property twin with 3D-consistent renovation editing and cost/ROI decision support for solo agents, FSBO sellers, and renovators — a direction the reports lay out but that is roadmap, not built. Today, HouseTwin is a capability showcase: flexible capture from cameras you already own, and an honest 3D preview.
Try it
Orbit a real reconstruction in your browser.
A photoreal 3D Gaussian-splat model rebuilt from 64 ordinary photographs — orbit, zoom, and pan, rendered live in WebGL2. Precomputed offline; no reconstruction runs in the browser. This is a placeholder capture that a front-garden model can replace.
Start here
Curious about 3D reconstruction from your own photos?
Whether it's a property, a site, or an asset, let's talk about what flexible-capture 3D can and can't do for your use case today.