Guided capture and the panorama viewer run for real, in your browser, right now.
3D World generation uses a fast on-device depth heuristic standing in for the cloud depth/diffusion model described in the build spec — swap in a real model in world3d.js / the backend worker without touching any other screen.
Everything lives in memory for this session. Nothing leaves your device — hook up backend/ for real accounts and persistence.
Walk the camera through 16 guided points, or drop in a panorama you already have to jump straight to the 3D World pipeline.
This takes about 5 minutes in production. Feel free to back out — we'll notify you the moment it's ready.