This page is about making the photo, not about the rules it has to satisfy. The full list of what a main image must and must not contain — the format list, the colour mode, the ban on text, badges and props — sits on the main image checker, along with what a browser can and cannot measure about a finished file. What follows is the seven-step route from an unlit product to a file that passes, with the number each step has to hit.
The reason it needs to be a sequence rather than a checklist is that the steps constrain each other. The frame shape you choose in step one decides whether the 85% fill target in step six is even reachable. The lighting choice in step two decides whether the edit in step five takes two minutes or eats the edge of your product. Doing them out of order is how people end up reshooting.
The frame fill requirement is usually read as an area measurement: the smallest rectangle containing your product, divided by the area of the whole image, should be at least 85%. That is how our checker measures it, and it is deliberately generous to awkward shapes.
It also has a consequence almost nobody states, and it is the single most useful thing on this page. A product's own proportions put a hard ceiling on how much of a square frame it can ever occupy.
Why a tall product cannot fill a square frame
Take a product whose long side is r times its short side, photographed square-on inside a square frame. Grow it until the long side touches the edges and it stops. Its bounding box is now the full frame in one direction and 1/r of it in the other, so the fill is 1 ÷ r and no crop will improve it.
A tall bottle at 3:1 tops out at 33%. A 2:1 box tops out at 50%. To reach 85% in a square frame the product has to be no more than about 1.18:1 — that is, within roughly a fifth of square itself.
Three ways out, in order of preference. Match the frame to the product and shoot 3:2 or 2:3 instead of 1:1. Turn the product diagonally, which is not a trick — a bottle laid across the diagonal genuinely occupies more of the thumbnail, which is what the rule is about. Or photograph it at a three-quarter angle so the silhouette squares up.
This is the step that decides everything downstream. One light pointed at a product on white paper lights both, and the paper always comes out darker than the product because it is further from the lamp and taking the light at a glancing angle. That is the origin of nearly every “my background looks grey” problem.
The standard fix is a second light aimed at the background alone, brought up until the background reads roughly one and a half to two stops brighter than the product. At that separation the background clips to white on its own, the product keeps its modelling, and the edit in step five becomes cosmetic rather than reconstructive. A white bounce card opposite the key light does the rest.
If you own a lightbox rather than lights, the same principle applies with less control: the product needs to sit forward of the back wall, not against it, so the wall gets its own share of light.
Cameras meter for an average, and a frame that is mostly white paper pulls the exposure down until the product is dark and the paper is grey — the worst of both. Meter or spot-check on the product, accept that the histogram will pile up against the right-hand edge, and check the highlight warning on the product only.
What you must not clip is product detail. A white mug on a white background is the hard case, and it is where the two-light setup earns its keep: the mug needs a visible tonal edge against the paper, which means it has to be at least slightly darker than the background it sits on. If it is not, no amount of editing will find an edge that was never recorded.
A well-lit background typically lands somewhere in the high 240s to low 250s, and that is a good result rather than a failure. Sellers on Amazon's own forums put it bluntly: pure white means RGB 255, 255, 255, and you will not get that in camera even shooting in a lightbox — it has to be finished in software.
Knowing this saves an hour of adding light until the product edge starts glowing. Once the background is clean, even and free of visible gradient, the camera's job is done. The remaining few levels are an editing operation, and doing it in software is more precise than doing it with lamps.
There are two routes and they fail differently.
The fast route is to pull the white point in a Levels or Curves adjustment until the background reads 255. It takes ten seconds and it is fine for a dark product with a hard edge. It is destructive for anything pale, because the same adjustment that pushes 250 to 255 also pushes the light side of your product to 255 — and the product loses its outline, its highlights flatten, and a cream ceramic becomes a silhouette with no top edge.
The safe route is to select the background and fill it. Any selection method works — a quick-selection tool, a subject mask, a pen path for a difficult edge — as long as the fill affects only the area outside the product. It is slower per photo and it does not touch a single pixel of your product. For a range you will shoot repeatedly it is worth doing once carefully and reusing the mask.
Whichever route you take, a cut-out product on pure white often looks like it is hovering. A soft contact shadow underneath fixes it and is not a prop; if you do not want to build one by hand, the shadow generator exists for exactly this gap.
Estimating fill by eye reliably overestimates it, because the eye judges the widest dimension while the requirement is about area. The arithmetic takes a moment and settles it.
The crop, worked out on a 2,000 px square
The frame is 2000 × 2000 = 4,000,000 pixels. Eighty-five per cent of that is 3,400,000 pixels of bounding box.
For a product that is roughly square, the box needs to be √3,400,000 ≈ 1,844 pixels on a side. That leaves (2000 − 1844) ÷ 2 = 78 pixels of white on each edge.
Seventy-eight pixels is a margin of under four per cent per side. It is far tighter than most people crop, and it is why so many technically compliant photos still measure in the sixties. If you want the number checked rather than calculated, drop the file into the main image checker and it draws both boxes.
Convert to sRGB and embed the profile. A file left in a wide-gamut working space renders differently in every browser that ignores the profile, which turns a colour complaint into a systematic one.
Save JPEG at high quality, and expect the flat white to move. JPEG compression perturbs large flat areas by a level or two, so a background you filled to exactly 255 can come back as 253 — which is why the checker offers a tolerance setting rather than demanding a literal 255, and why you should re-measure the exported file rather than the one on your canvas.
Size it generously on the long side. Zoom is the reason resolution matters, and enlarging a small file cannot add detail that was never captured; upscalers invent texture your product does not have, which on a fabric or a wood grain is worse than no zoom at all. If you need the same shot in several marketplace shapes, the resizer pads and re-squares a batch in one pass, and note that a listing image has no physical size — the reason a DPI figure means nothing on screen and everything in print is worth two minutes if your export dialogue keeps asking for one.
The three whites, and how to tell which one you have
Everything above is really an argument that “white background” describes three different states that look similar in a thumbnail and behave completely differently.
Shot on white card with no dedicated background light. Looks white in isolation, reads as a faintly grey card the moment it sits in a results grid next to a genuine 255 neighbour. Usually drifts across the frame — clean in the corners, dirtier halfway up an edge where the light fell off.
The fast Levels route applied to a pale product. The measurement passes and the photograph is worse: the top edge of the product has merged into the background, highlights have gone flat, and shape information the buyer needed is gone. This is the failure that a background checker cannot see, because by every number it is a pass.
Selected and filled, product untouched. The only one of the three that is both compliant and honest. Costs a minute more per photo than the blown version and is the difference between a listing that looks professional and one that looks like it was rescued.
None of this transfers wholesale to other marketplaces, which measure different things — Etsy works to a shortest-side figure and crops the first photo to a different ratio in search, set out in the Etsy image size and crop reference. And if the same artwork also goes to a print partner, the resolution it needs on the garment is a separate and much larger constraint, covered in the print-on-demand image requirements.
Written 2026-07-28. The photographic and arithmetic content here is not marketplace policy and does not expire; the rules it is aimed at satisfying do, and are not restated on this page.
The RGB 255, 255, 255 requirement for a main image, the 85% frame-fill figure and the resolution guidance are Amazon's, discussed by sellers and Amazon staff in the Seller Central forums — see the pure-white background discussion, which is also the source of the point that a camera will not produce 255 even in a lightbox, and Amazon's Listings Lounge thread on image requirements. Amazon's own image documentation sits behind a Seller Central sign-in and varies by marketplace and category; confirm the current requirements there before reshooting a range. The one-and-a-half to two stop background separation is standard studio practice rather than a marketplace rule. The fill arithmetic is our own, based on measuring fill as a bounding box over frame area — the method used by our main image checker; a marketplace measuring silhouette area instead would produce lower numbers for irregular products.