Furnish a scanned room

Pick a piece from the palette, point at the floor of a real scanned room to place it, and see its distance to the wall and to its neighbours.

A capture of the project running in Graspable's emulated headset. A scanned seating area seen from the headset: an oak dining table with two chairs and a terracotta armchair standing on the concrete floor in front of a dark sofa and a yellow staircase, an oak bookshelf against the far wall, a red footprint under the armchair with plates reading 13.8 cm and 37.8 cm to its neighbours and 359.5 cm to wall above it, a controller pointing at the armchair.

Open the appContinue from it in GraspableRuns in a browser, on a Quest or a Vision Pro.

The request, word for word

This is the text that was sent to the agent to start the project. Nothing was edited for this page.

Let me furnish this room. Replace the starter side table with a planner for the open floor in front of the rug. A palette floats at my left, at chest height, with five pieces: an oak dining table (160 x 90 cm, 75 high), a dining chair, an armchair in terracotta fabric, a tall oak bookshelf (80 x 30 cm, 190 high) and a brass floor lamp with a linen shade that really glows. Pointing at a piece in the palette and pulling the trigger (click on the desktop) puts a new one on the floor in front of me. I move a piece by pointing at it, holding the trigger and dragging it across the real floor; it stays on the floor, and the thumbstick left/right (Q and E on the desktop) turns it in 15 degree steps. The B button (Delete on the desktop) removes the piece I am pointing at. While I move a piece, show its distance to the nearest wall in centimetres on a small plate above it, and draw its footprint on the floor. When two pieces are less than 60 cm apart, the gap between them shows a red line with the distance, and the two footprints turn red; otherwise footprints are a quiet white. Start with a small arrangement already in place so the room does not look empty: the table with two chairs, and the bookshelf against the left wall. The furniture has to look real next to the photo scan: built from proper parts (legs, table top with rounded edges, cushions, shelves with a few books), PBR materials with wood grain and fabric textures drawn in code, lit to match the room's ceiling lamps, with soft contact shadows on the floor. The palette shows a small model and the name and size of each piece on dark plates with light text that I can read from two metres. It has to work in VR in the emulated headset with the controllers and on the desktop with the mouse. Keep it fast: the room scan is heavy.

How it went

Runs
3 runs: the request above, then 2 follow-up requests. A run ends only when the project builds and loads in a real browser without errors.
Started from
The Scanned Room template: Start inside a 3D capture of a real room, where what you add sits on the real floor, walls and furniture at real scale.
Notes and credits
Room capture from the Eyeful Tower dataset. Copyright (c) Meta Platforms, Inc. and affiliates. MIT License. The first build had flat slabs for furniture; the second request rebuilt the five pieces from real parts, the third fixed the clearance check, the thumbstick and the B button. A held piece follows in steps, not smoothly, and cannot be pushed straight away from you, because the piece itself blocks the pointer. AI: Graspable agent (build), Claude Code (both follow-ups).

Build your own

Graspable is a desktop app. Describe the scene you want, and the agent builds it as a real project on your computer and checks that it runs. It is free with your own AI account.

Download GraspableYour first project

More examples

See all examples