If you sell on Amazon or any large marketplace, your background is not a design choice. It is a rule. Get it wrong and the listing gets suppressed, or worse, the image does not get picked for the thumbnail. This gets tricky when you are generating or editing product shots, because the tool is happy to give you a beautiful scene that a reviewer will reject in four seconds.
I have been through this loop a few times with sellers who use the AI image generator on our homepage to mock up product shots before paying for a real photoshoot. Here is what I have learned about backgrounds, in the order that actually matters.
Start with the rule, not the picture
A typical main image needs a pure white background (RGB 255,255,255), the product filling most of the frame, no props, no text overlays, no watermarks, no borders. That is the compliance layer. Everything after that is art direction.
So the first thing to check when a shot comes back is not whether it looks nice. It is whether the white is actually white. Generated images often give you a slightly creamy background. Screenshot it, open it in any editor, and sample the corner pixel. If it reads 250,250,248, you are not compliant yet, even though nobody would notice with the naked eye.
Generate the product, composite the background
The mistake I see most often is asking for the whole thing at once: product, shadow, white studio, done. You end up with a shadow that eats into the product edge, or a faint gradient that looks like a studio but fails the checker.
A better pipeline is two steps. First, generate or clean the product on a simple dark or mid-grey background where edges are easy to see. Then cut it out and drop it on true white with a soft contact shadow you control. That second step takes ten minutes and saves you a rejected listing.
Secondary images are where you can stretch
Main image is strict. Positions two through seven are where you get to tell the story: lifestyle context, scale reference, detail crop, packaging. This is where a generated scene is genuinely useful, because you can put the product on a kitchen counter or a hiking trail without booking either location.
Two things to keep in mind. One, the product itself must not change between shots. If the label shifts colour or the stitching moves, buyers notice and trust drops. Two, keep the same light direction across the whole set. Mixed shadows read as fake even when each individual shot is fine.
Contrast and the thumbnail problem
Most buyers first see your image at about 180 pixels wide. At that size, a pale product on white disappears. Before you export, shrink the image down and look at it on your phone at arm's length. If the silhouette is not clear, darken the product edges slightly or shift the product to fill more of the frame. This is a lighting and framing fix, not a background fix, but it is the single highest-value tweak for a marketplace listing.
A quick checklist before upload
- Corner pixel is 255,255,255
- Product occupies roughly 85 percent of the frame on the main image
- No text, badges, or props on the main image
- Shadow is soft, falls the same direction as every other image in the set
- Product colour matches the physical item under neutral light
- Image is at least 1000 pixels on the long edge so zoom works
Where generation helps, and where it does not
Generation is great for concepting, for filling gaps in a set, for testing a packaging change before you print, and for lifestyle scenes you cannot afford to shoot. It is not great for the final main image of a product where the exact printed label matters, because small text and colour drift are still unreliable. Use it to plan and to fill, then photograph the hero shot.
If you want a starting point, the prompt pattern is boring on purpose: product name, material, size, light direction, background type, camera angle. Boring prompts hold together. That is the whole trick, and it is the same one covered across the generator's guides if you want more on keeping a set consistent.