Looking for direction: AI before/after interior impressions for real estate
Hi All, Working on a project which could be my first paying customer. I could use some outside experience on. The idea: take a real photo of a room and turn it into a clean before/after impression to help sell a property. So first stripping the room back to a blank canvas. Clearing the clutter, removing the furniture, sometimes even taking out a wall to show what the space could become. And then dressing it back up the way an interior designer would, so a buyer can actually picture living there. The part I keep running into is proportions. The room has to stay true to its real dimensions. A wall that quietly shifts, a ceiling that grows a bit, a window that drifts. For real estate that breaks the whole thing. Ideally I'd work from the floor plan so the perspective stays locked and nothing gets invented. The setup so far leans on Claude Code to drive the workflow and write the edit prompts, plus image tools like Higgsfield for the actual generation. Open to other tools if something works better. A few questions for anyone who has been down this road: • Has anyone built something like this, real estate or interior, and managed to keep it dimensionally accurate? • What's your approach for stripping a room and reconstructing the space behind a removed wall without the model hallucinating something that was never there? • Any tools, workflows or tutorials you'd point me to? Higgsfield, Nano Banana, Flux Kontext, ControlNet, floor-plan-to-3D, whatever has actually held up for you. Happy to share back what I figure out. Where would you start?