Step 1
Open a room photo
Empty room or viewport still. Camera and walls become the locked base.
Interior AI · precise 3D placement
When the client says “move the chair left,” the whole room shouldn’t reshuffle.
Prompt-based tools like GPT Image invent furniture into a photo — fine until a small layout note forces a full re-generation. Pointlight places real 3D models on the room photo first. Position and scale stay under your control. Then you render.
Process
Empty room → place real 3D furniture on the photo. You control position and scale before the photoreal pass.
Result
Three cuts of the finished room
Same layout after render — wide, detail, and material cuts. Placement decided the composition; the render finished the look.
Why prompts lose the layout
Most AI staging regenerates the whole frame. Asking for “chair a bit left” often re-rolls the sofa, the rug, and the lamp — because the model never stored furniture as movable objects.
Pointlight’s placement step is spatial: grey blockout models sit on the photo with controllable position and scale. A client note becomes a drag, not a new lottery ticket.
We do not claim GPT Image is useless for first-pass beauty. We claim a different job: keep the layout you decided while you iterate.
Pointlight vs prompt-based staging
| Pointlight | GPT Image / prompt staging | |
|---|---|---|
| How furniture enters the photo | Real 3D models you place | Model invents furniture from text |
| “Move the chair left” | Drag that object; others stay | Often a full re-generation |
| What you lock before beauty | Layout, scale, grounding | Hope the next sample matches |
| Best for | Client iterations on placement | One-shot mood boards |
Also in Pointlight (separate workflows, not this page): region tags for fast local edits, and building a room shell when you need geometry from scratch. For multi-turn cleanup after layout is locked, see HD redraw / revision stability.
How to place 3D furniture on a room photo
Four steps. Layout first, beauty second.
Step 2
Drop 3D models
Pull furniture from the library onto the photo. Grey blockout shows exact placement.
Step 3
Adjust one piece at a time
Move, scale, rotate. Client notes become object edits — not a new global prompt.
Step 4
Render
Photoreal pass on the locked layout. Same positions, finished materials and light.
FAQ
How is this different from GPT Image for furniture placement?
GPT Image invents furniture into a photo. Small layout notes often force a full re-generation. Pointlight places real 3D models first — you control position and scale — then renders. Layout edits are object moves.
Can I move one piece without reshuffling the room?
Yes. Select the 3D model and drag it. Other objects keep their positions because placement is scene-based.
Do I need SketchUp or Blender?
No. Start from a room photo in the browser. Other 3D screenshots can be a base, but they are optional.
What do the grey models mean?
They are precise 3D placements before the photoreal render — scale, grounding, and spacing you can still change.
Is this virtual staging for real estate only?
No. Designers, staging pros, and anyone iterating room layouts on photos use the same placement → render loop.
Place the sofa where it belongs — then render.
Free trial credits. No card required to start.