Guide
How to photograph a room so AI design actually works
Last updated: 14 October 2026
Everything an AI room designer knows about your space comes from one photograph. Five minutes of care taking it changes the quality of every design you get back.
Stand in a corner
This is the single highest impact thing you can do. From a corner, the camera sees two walls meeting and a run of floor, which is exactly the information a model needs to understand the shape and size of the space.
From the middle of a room, or flat against one wall, you get a shallow view with no depth cues. Designs generated from that kind of photo tend to come back with furniture at odd scale, because the geometry was ambiguous.
Include the floor
Point the camera slightly down so a meaningful amount of floor is in frame. The floor is where furniture goes, and it is also the surface that tells the model how big everything else is.
A photo composed around the ceiling or the top half of the walls gives a model very little to work with.
Use daylight, and do not shoot into the window
Daytime, curtains open, lights off is usually the best combination. Mixed lighting, where a warm bulb fights cool daylight, produces colour casts that follow through into the design.
Avoid pointing the camera straight at a bright window. The camera exposes for the window, the room goes dark, and detail the model needs disappears into shadow.
Hold the phone straight
Keep the phone roughly vertical rather than tilted up or down, and hold it at about chest height. Tilting introduces strong perspective distortion, which makes walls lean and confuses the sense of scale.
Most phone cameras have a grid or level option. Turning it on for one photo is worth it.
Five mistakes that ruin the result
- Shooting from the doorway so only one wall is visible.
- Standing too close to a large piece of furniture, which then dominates the geometry.
- Shooting at night under a single warm bulb.
- Holding the phone at eye level and tilting down, which compresses the floor.
- Cropping the photo afterwards, which throws away the edges the model uses for context.
What to do with the result
Generate more than once. The first output is a starting point, and the second or third attempt with a slightly different photo or instruction is often noticeably better.
Then treat the picture as a decision, not a specification. Before buying anything in it, measure the piece, measure your space and measure the route in.
A good photo takes about thirty seconds and improves every design you generate from it afterwards. It is the highest return half minute in the whole process.
Common questions
Do I need to tidy the room first?
No. A lived in room gives the model more information about the space than an empty one does. Clear anything you would not want stored on a server, but you do not need to stage it.
Do I need a wide angle lens or a special camera?
No. A normal phone camera is fine. Standing in a corner does more for the result than any lens.
Why did my design come back with the wrong proportions?
Usually because the photo was taken from the middle of the room or from very close to one wall, which gives the model too little geometry to read. Back into a corner and include the floor and two walls.