A compositing technique for product photography: rather than fully deleting an AI image generator's own placeholder rendition of the product and replacing it with a cutout of the real product photo, selectively preserve the AI-rendered shadows and contact points (where the product touches a prop — resting on fruit, sitting on a ledge) and paint the real product cutout in around them.
This works because modern AI image models (ChatGPT/GPT-4o cited as the example) can render a placeholder product that is already ~95% visually accurate, including plausible contact shadows the model inferred from the scene it generated. A straight copy-paste of the real product cutout onto the AI background discards that contact information and can read as pasted-on; erasing only the non-contact regions of the AI placeholder and letting the real cutout occupy the rest keeps the physically-grounded shadow/contact detail.
Part of the finishing stage of AI-Background-First Product Photography Workflow; complements a whole-image grain pass and Neural Filters Harmonization (PNG-to-PNG) for final cohesion.