AI product photos on a white background for Amazon, Etsy and other marketplaces

Search for a white background product photo and you will find two kinds of advice: tools that promise a perfect listing image in one click, and forum threads about a listing that was suppressed over a main image that looked white and was not. This post sits between them. It covers what marketplaces such as Amazon and Etsy typically ask of a main image, how to steer an AI studio toward it, and which parts no model can promise.

What a main image is usually asked to be

Rules differ by marketplace and change over time, so read your own seller guidelines before you build a batch. The common core is short enough to remember, though.

• A pure white ground. Not cream, not light grey, not white fading to grey in the corners. Many marketplaces want it to read as true white.
• The product fills the frame. Large and centred, with a thin margin, rather than a small object floating in empty space.
• Only what is for sale. No props, no lifestyle scene, no badge, no watermark, no added text or logo overlay.
• A real depiction. The picture must show the product the buyer will receive, in its true colour and shape.
• Enough pixels. A listing image has to stay sharp when a buyer zooms, so size matters more here than on a social post.

The first item is the one AI pictures miss most often, and the last item is the one most people forget. Both come up again below.

Step 1: ask for the studio, not just the colour

Typing "on a white background" gets you a background that is mostly white. The model has seen millions of photos with a white wall, and a white wall is lit unevenly, so the grey creeps in toward the edges. A better prompt describes the setup that makes white in a real studio: a seamless sweep, even light from both sides, and no vignette.

Keep it to one product per frame, one angle, and say the angle in words. For lighting vocabulary that carries over, see the product photography guide, and for shiny or transparent items read the posts on chrome and metal and glass and liquid first: they change what "white" has to reflect.

Step 2: choose a frame shape that lets the product fill it

A product that fills the frame is partly a composition question and partly a canvas question. Many marketplaces show main images in a square tile, and a square canvas lets a mug, a bottle or a box fill the picture without a wide strip of empty white on each side. Tall items such as a lamp or a bottle suit a square or a slightly tall frame; a long item such as a belt or a skateboard may need a wide one. Set the aspect ratio in the model settings before you generate rather than cropping later, and the aspect ratios guide explains how the options behave.

Step 3: if you already have a photo, edit it, do not regenerate it

A generated mug is a picture of a mug that looks like yours. If a buyer will receive your actual mug, start from your own photo and change only the background. An edit model takes the photo plus a sentence, and the sentence has two halves: what to change, and what to keep.

This is the same split described in the post on background removal or an edit prompt: a removal model leaves an empty layer, an edit model paints a new backdrop. For a white listing image either works. A cut-out placed on white in your own layout gives you the exact white you choose, while an edit model gives you a shadow and a lighting match in one step. Wording for the change and keep halves is in the guide to writing edit prompts.

Step 4: fix the white you got

If the first result is close but a little grey, do not start over. Send it back to an edit model with a narrow instruction: make the background pure white, brighten it evenly to the edges, leave the product and its shadow alone. Do this before any other tweak, because every later pass can nudge the tone again. If a tint remains, the reliable fix is outside the studio: open the picture in any image editor and push the white level up on the background only. A quick way to see the problem is to place the image on a pure white web page and look for the edge of the picture.

Step 5: check the file at full size

Open the final image at full size, not as a thumbnail, and look for the failures a thumbnail hides:

• Grey or cream patches in the corners, and a visible rectangle where the edit stopped.
• A halo of the old background along fine edges such as hair, fabric fringe or wire.
• Printed text on the product that has become soft, wrong or invented, and logos that have drifted.
• A changed colour, an extra handle, a missing button: the picture must match the item you ship.
• A shadow that reads as a dirty smudge, or a floor line that has appeared.

If the product needs more pixels, run an upscaler on the final image and check it again, since an upscaler can change fine details. The upscaler post explains when to keep it faithful.

What the studio cannot do for you

The studio does not know any marketplace's current rules, does not read your listing, and cannot guarantee that an image will be accepted. Required pixel sizes, how much of the frame the product must fill, whether a shadow is allowed, whether an AI-made or AI-edited image needs a disclosure, and what the category allows are all set by the marketplace and can change, so check its seller help pages for your category before you upload. A generated product is also not your product: for items where the exact look matters, such as apparel colours, a handmade piece or a label with regulated text, edit your own photo and compare the result against the real item. For a wider view of images versus short clips on a product page, see image or video for a product page, and the models page lists the edit and image models you can compare on the same photo.