World models are learning real physics.
We're building the other half: the real things you already own, as editable 3D you can place and test before you build.
“Will it fit?” is becoming a question you should be able to answer before moving a single thing.
And once the room is editable in 3D, you can keep going: move furniture, test layouts, see what works.
Me: Generate a cozy living room for me.
@NanoBanana: Here you go.
Me: Looks great. Now let me rearrange the furniture and buy the actual pieces.
Aroomy:
Okay, this part is kinda nice!
Put the acrylic shelf into Aroomy, played around with the placement, found the setup I liked, then hit Buy.
@OpenAI when does the “generate a room” part turn into “buy the furniture” ? 👀
#SpatialAI #3DAI #InteriorDesign #Aroomy
Seeing space ≠ being able to use space.
Cool demo - pulling walls and furniture out of a video. But until that data survives alignment, scale correction and real world execution, it’s just a labeled point cloud.
The real challenge is the gap between spatial detection and spatial editing.
a tiny AI 0.5B just turned a normal phone video into a fully labeled 3D room run phone.
SpatialLM that understands walls, doors, furniture, and everything layout.
it trained as a standard multimodal LLM.
Most vision models see pixels, SpatialLM sees space.
Open models : github.com/manycore-research…
Stop paying the “trial-and-error tax” on home decor.
Look here: with aroomy, you can auto-retrieve & segment, change colors, modify panels, and edit layouts on the fly.
Best part?
Every object is connected to a real-world product you can buy.
Try aroomy now👇🏻
aroomy.com/
3D creation is more useful when it connects to real products.
Aroomy allows you to search through products from real stores, turn them into 3D with one click and drop them directly into your space.
The useful part is not just to convert an image to 3d.
It’s bridging 3D to real world things you can actually buy.
#3DAI #SpatialAI #ComputerVision
This is where Spatial AI gets interesting.
Not just generating a 3D scene — understanding it well enough to keep editing it.
Objects, layouts, dimensions, relationships.
Generation is the start. Editability is the workflow.
This is the kind of AI coding demo I want more of.
Give it a floor plan. Turn it into a 3D room. Then start messing with everything.
Move the bed. Change the dimensions. Rearrange the furniture. Rotate the entire room. And it updates fast enough that you can keep iterating without constantly waiting on the model.
I built this with Space Bunny in OpenCode mostly to see how it handles something spatial instead of another normal web app.
Way more interesting.
The model has to understand the layout, keep track of where everything belongs, and still make the room visually coherent while you keep changing it.
AI coding gets a lot more fun when you stop asking it to build software that already exists.
Opus 5.5 looks incredibly smooth. The lighting and material swaps are top-tier.
But I tried it. You're locked inside a pre-defined room. You can move the sofa. You can't raise the walls.
At Aroomy, we took a different route. Start from an infinite canvas. Build your own kingdom from scratch.
Bring your products into 3D.
Aroomy helps brands turn products into interactive 3D experiences — giving customers a new way to visualize, customize, and explore before they buy.
Featuring "felt right" an official affiliate.
#ecommerce #3DCommerce #SpatialAI






