AI Space Visualizer
Generative visual commerce for furniture retailers
A shopper photographs their room and sees real, in-stock furniture placed in it — the retailer's actual catalogue, not lookalikes. They can rearrange the result by describing the change, then send the design to the retailer's sales team.
studio.tomyest.comThe problem
Furniture is bought on imagination. A shopper cannot tell whether a sofa suits their living room until it is in their living room, and that uncertainty is where the sale is lost — or comes back as a return.
What made it hard
Generative image models will happily produce a beautiful sofa that the retailer does not sell. The system has to place products the catalogue actually stocks, in the sizes it stocks them, in a room whose geometry it inferred from one photograph — and it has to be honest when the catalogue has nothing that fits.
What we built
Catalogue products are matched to positions the room can actually take, then composed into the photograph. The result is a before-and-after the shopper can drag between, revise in free text, share as a link, and hand to the retailer's sales team as a finished design with the products named.
Architecture
- Room analysis from a single uploaded photograph
- Placement constrained to catalogue products the position can take
- Generative composition into the original room image
- Rearrangement by free-text instruction rather than a layout editor
- Signed share links, and a hand-off that reaches the retailer's sales team
The product
What became possible
- A shopper sees the retailer's real, in-stock products in their own room, and the before-and-after is one drag apart.
- Revisions are described rather than drawn — no layout editor to learn.
- A finished design reaches the retailer's sales team with the products already identified, so the conversation starts from a specific room.