Skip to content
GPU-assistedUpload required

On Your Own Store, Nobody Enforces the Rules but You

A marketplace suppresses a listing that breaks its image rules, which is annoying and also a favor — the standard is decided for you and enforced for free. On your own store, nothing is enforced. Every image you upload appears exactly as supplied, and the result is usually a collection page where one product is large and centerd on white, the next is small on a gray wall, and a third is a phone photo on a kitchen counter. No single image is bad. The grid is what customers actually look at, and the grid looks like a folder.

Use this without the search next time. Prathom Workbench puts Prathom's tools in your toolbar.

Add to Chrome — free

Drop an image here, or click to browse

Up to 25 MB. Uploaded temporarily, deleted after processing.

GPU-assisted — results vary slightly between runs.

What it does

  • Background replaced with white
  • Every output normalized to about two megapixels
  • Light sharpening after the resample
  • No generative model, product pixels untouched

How to use Shopify Product Images

  1. 1

    Decide your standard before you process anything

    One background, one framing convention, one aspect ratio, one file naming pattern. Writing it down takes five minutes and is the entire difference between a catalog and a folder.

  2. 2

    Run every image through the same pass

    The product is isolated onto white and normalized to a common pixel count, then lightly sharpened to recover what the resample softened.

  3. 3

    Review the collection page, not the images

    Open the grid and look at it as a customer would. Mismatched scale, background, and crop are only visible there, and that is where they cost you.

How it works

Three deterministic steps and no generative model.

The product is matted with a background removal model and composited onto a white plate at the original dimensions, which handles the background.

The result is scaled to about two megapixels, which handles size. This is the step that makes a set consistent — a twelve megapixel photograph and a two megapixel one end up at the same pixel count, and the platform's thumbnails are then generated from equivalent sources.

A light sharpening pass runs last, because any resample softens edges slightly and a small amount of edge contrast puts the apparent crispness back.

Nothing samples randomly, so the same input always produces the same output.

Consistency is the only thing the grid shows

A customer on a collection page is not evaluating photographs. They are scanning a grid, and what a grid makes visible is difference.

If one product is shot closer than another, the smaller one reads as a smaller product, whether or not it is. If one background is white and another is warm gray, the warm one reads as older stock. If one image is sharp and the next is soft, the soft one reads as lower quality.

None of those readings are deliberate, and none of them are about the products. They are artifacts of the images having been made at different times under different conditions, and they are invisible while you are looking at any single photograph.

Which is why the review that matters is the collection page rather than the image. Open it, scroll it, and look for the one that stands out. That one is either your best photograph or your worst, and either way it is the one that needs to change.

What a standard actually contains

Writing it down is the part people skip, and the list is short.

Background: one color, stated as a value rather than as a word.

Framing: how much margin around the product, expressed as a proportion of the frame so it applies to products of different shapes.

Aspect ratio: one, usually square, chosen because it crops predictably in every theme and every thumbnail.

Size and format: a pixel count and a file type, so weight is predictable.

Naming: a pattern with the SKU in it, because in eighteen months somebody will be trying to find the source file for a product they did not photograph.

Every item on that list is a decision you make once. The value is not that any particular answer is right; it is that the twentieth product gets the same answer as the first.

Where this page stops

Two things it does not do, both of which matter for a real store.

It does not crop. Framing consistency is a crop decision, and a crop cannot be applied blindly across products of different proportions without cutting some of them. Use the crop tool, per product, against your written framing rule.

And it does not handle the images that are not product shots. Lifestyle photography, scale references, detail crops, and packaging shots all belong on a product page and none of them should be run through this. They live under different rules, and mixing them into the same treatment is how a store ends up with a lifestyle photograph awkwardly cut out onto white.

Publication gate

This page ships once the workflow has been run against a set of at least six products photographed in different sessions, plus a single already-white hero shot, with the outputs viewed as a grid, their pixel counts confirmed equal, and the sharpening checked for halos at full resolution.

Examples

A range shot over several sessions

products/*.jpg - mixed sizes, mixed backgrounds, mixed cameras
prathom-shopify-product-images.png - about 2 MP each, white background

The intended case. Background and pixel count converge exactly, which removes two of the three things that make a grid look assembled.

A single hero shot

hero.jpg - 4000x3000, already on white
hero-normalized.png - about 2 MP

Little to gain. There is no grid to be consistent with, and normalizing a large image downward loses detail that the zoom view would have used.

Frequently asked questions

What size should product images be?

Large enough that the zoom view is worth having and small enough that the page stays fast. Around two megapixels is a reasonable middle for most stores, and the platform will generate the smaller thumbnails from it. The important part is that every image is the same, not the specific number.

Does image weight really matter?

On your own store it is your bill and your load time. A collection page shows dozens of images at once, and their combined weight is usually the single largest factor in how quickly the page becomes usable. This matters more here than on a marketplace, where the platform resizes aggressively on your behalf.

Should the background always be white?

Not necessarily, but it should always be the same. White is the safe default because it disappears against most store themes and survives dark mode reasonably. A consistent light gray or a consistent brand tone is equally valid; a different background per product is what fails.

How is this different from the marketplace page?

That one exists to satisfy a rule you did not write, so it produces pure white and nothing else. This one exists to impose a rule you did write, so it also normalizes size and sharpens, which are choices rather than requirements.

Is the AI Shopify Product Images tool free, and do I need an account?

It is free and there is nothing to register for. No account is created, no address is collected, and no reduced-quality preview is withheld for a plan that does not exist. The constraints that do apply come from the workflow itself rather than from billing: 25 MB in, a shared GPU queue, and a thirty-minute window to collect the output.