Put a Photo on a White Background

How it works

Four steps, and the subject is never touched.

A segmentation model reads the photo and produces a mask marking subject from background. The dimensions of your original are measured. A canvas of pure white is created at exactly those dimensions. The subject pixels are composited onto it through the mask, without rescaling.

Only the pixels outside the mask change. Color, exposure, and detail inside it are the ones your camera recorded, which means a label stays readable and a product color stays accurate — and equally, any problem in the original photo comes through untouched.

Why pure white and not nearly white

The fill is 255, 255, 255 exactly.

That specificity exists because marketplaces specify it. Amazon's main image requirement names pure white, and compliance checks are automated — they sample pixels and compare values. A backdrop at 252 looks identical to a human and is a different number to a script.

Anything with a color cast, however faint, also becomes obvious the moment your image sits in a grid beside other listings on an actual white page. One slightly warm background among twenty neutral ones reads as a mistake.

The pale subject problem

This is the one case where the tool does its job correctly and the result is still wrong.

An outline is visible because of a difference in value between an object and what is behind it. Put a white shoe on a white field and there is no difference along most of its boundary. The parts of the shoe that carry shading — the sole, the creases, the shadowed side — still read. The bright toe cap does not. The product appears to dissolve at the edges.

No mask setting changes this, because the mask is correct. The subject really was separated; it simply cannot be seen against the color it was placed on.

The answer is a backdrop that is not white. A very light gray — around 242 on all three channels — still reads as neutral in a page layout, still looks clean in a listing grid, and leaves enough separation for a white product to keep its outline. The product background tool on this site is built around exactly that value, and for a pale subject it is the right page.

Choosing a source photo

Contrast is what the separation step needs, and it needs it between subject and background rather than within the subject.

A dark object on a light surface, or a light object on a dark surface, both work. The failure is same-on-same: a black camera on a black cloth, a white mug on a white table. There the model has no boundary to find and will either take part of the object or leave part of the surface.

Thin structures are the second thing to check. Handles, straps, stems, cables and chair legs occupy few pixels and are the first thing a mask drops.

Publication gate

This page ships once the workflow has been run against a dark product on a light surface, a white product, and an object with thin structures, with the outline and any surviving background fragments inspected at full size.