Amazon image compliance, checked across the whole catalogue

Amazon’s main-image rules are precise and published: pure white background, product filling most of the frame, no props, no watermarks, minimum dimensions. They are also enforced quietly. A non-compliant image does not usually produce an error you notice — the listing simply stops being shown, and you find out from the sales figures.

Checking whether an image meets those rules is arithmetic on pixels. Which makes it exactly the kind of job to hand to a machine, and exactly the kind of job where “looks fine to me” is not an answer.

Every image gets a verdict, not a score

A catalogue scan puts each image into one of a small number of buckets, and the bucket tells you what to do:

• Already compliant — nothing to do.
• Cleanable — the background is nearly white and can be corrected without touching the product.
• Too small — below Amazon’s minimum, and enlarging it would only produce a blurry rejection.
• Needs reshooting — a lifestyle shot, a busy background, a product too small in frame. No amount of processing turns this into a main image.

That last category is the useful one. Most tools try to fix everything and quietly make some things worse; being told plainly that eleven products need photographing again is more actionable than eleven silently degraded images.

Why it prepares fixes rather than applying them

Automated background removal is genuinely dangerous on a real catalogue. A flood fill that walks in from the edges will keep going wherever the colours are close enough — and on one run here, that meant it walked through a pale background, into a model’s face, and erased it. The image was technically compliant afterwards. It was also unusable, and nothing in the process knew that.

So the tool reports what is wrong and prepares the corrected version for review. Applying a change to your image library is a deliberate act, taken after somebody has looked.

Transparency is not white

A worked example of why this needs a real check rather than a glance: a PNG with a transparent background looks perfectly white in every previewer you own, because the previewer renders transparency as white. Amazon does not — it discards the alpha channel, and what you get is whatever sits underneath, which is frequently black.

That is a listing that looks correct in every tool you have and is suppressed the moment it arrives. It is checked for explicitly.

Scope it before you run it

A scan can be limited to one brand, one SKU or one product type, which matters when you are working through a backlog rather than doing everything at once. Start with the suppressed listings, confirm the verdicts match what Amazon is telling you, then widen it.

What this does not do

It does not photograph anything, and it will not rescue a bad photograph. It tells you precisely which images are costing you listings and which of those are worth a machine’s attention — the reshoot list is still a list of things somebody has to do properly.