We are an AI studio, so you might expect this article to tell you AI can do everything. It cannot. After thousands of delivered frames for UK home, garden and DIY brands, we know precisely where the technology is genuinely production ready, where it needs a craftsman holding the reins, and where we refuse to use it at all. Here is the honest map.
Where AI is genuinely production ready
- Lifestyle and room scenes. Your product in the garden, kitchen or hallway it was made for, in any season and any light. This is the category AI has completely changed: scenes that needed a location, a crew and a weather window now come from a single supplied photograph.
- Range consistency at scale. Forty SKUs in the same room, same light, same styling. A physical shoot drifts across days; a production rig holds the line across hundreds of frames. Large ranges are where AI is strongest, not weakest.
- Seasonal refresh. The same range re-dressed for spring launch, summer promotion and Christmas trading, without re-shooting anything.
- Hero-on-white conversion. Turning warehouse photography and old pack shots into clean marketplace-grade heroes.
Where it fails without craft
- On-pack text and labels. Generative models garble small print. The professional approach reproduces packs from the real artwork and checks every character, or keeps the label photographic rather than generated.
- Material truth. Flooring is the classic: models love inventing grout lines on click vinyl and erasing the bevel that tells a buyer what the product actually is. It takes category knowledge to even notice, which is why generalist AI output fails retailer checks that specialist output passes.
- Scale and proportion. Left alone, models beautify: handles grow, planters shrink, ladders stretch. Every frame has to be checked against the physical product's true dimensions.
- Logos and brand marks. Never described in words, always locked from the real artwork. A nearly-right logo is worse than no logo.
Where we refuse outright
Safety-critical detail does not get invented in this studio. If a ladder's tread pattern, a fixing specification or a load-bearing feature has never been photographed, we do not imagine it, we ask for one real photograph and build from that. An image that misrepresents a safety feature is not a creative decision, it is a liability, and no deadline is worth it.
The reference is the product. If a detail is not in the reference, it does not go in the frame.
How to brief AI like a photographer
- Send the real thing. Listing link, physical product photography from several angles, pack front and back. Reference quality decides output quality, nothing else comes close.
- Name the room, the light and the audience. "Bright contemporary British kitchen, morning light, aimed at trade buyers" beats three paragraphs of adjectives.
- Give the mood in pictures. A three-image mood board communicates more than any written description, and a good studio will extract palette, materials and furniture language from it for every frame.
- Fix everything in one round. The best results come from one consolidated amend list against the first pass, not five rounds of drift. Iterating endlessly on top of AI images degrades them; professionals rebuild from the original instead.
Put the finished frame next to the physical product and the retailer's spec sheet. If anything on the image could not survive that comparison, it does not ship. That is the whole discipline, and it is the difference between AI imagery that wins listings and AI imagery that gets bounced.