Filling the Frame Equally Makes Every Product the Same Size

How it works

Each photograph is scaled to about one and a fifth megapixels and repainted at a moderate denoise against one fixed description that does not vary by product: a plain light background with a gentle falloff, one soft key from the upper left, the product standing on a neutral surface with a soft contact shadow, and the same camera height and distance every time.

Two clauses do the work specific to this page. The product must be kept exactly as photographed, and it must not be resized within the frame.

The negative prompt names a resized product and changed framing first, ahead of the usual product-fidelity failures, because those are how a set stops being a set.

Scale is information, and it is easy to destroy

A customer looking at a product image is asking two questions before any other: what is it, and how big is it. The second one is answered almost entirely by comparison.

Within a range, that comparison is against the other products. If a 30ml bottle and a 500ml bottle are both photographed to fill their frames, they arrive as the same object at two labels, and the only remaining scale information is the number in the description, which people do not read carefully and cannot picture anyway.

The fix is at the camera. Photograph the whole range from one position with one lens, and let the small products be small in frame. Everything downstream — this page included — can then preserve a relationship that actually exists.

If the sources were shot individually, the relationship is gone and no processing recovers it. What you can do is rescale deliberately: take the real width of one product, work out the others in proportion, and resize each image so those proportions hold. That is arithmetic and it is reliable. Doing it by eye is not.

What converges and what does not

Being specific is more useful than promising a matched set.

Framing and camera height converge well, because they are geometric instructions with little room to interpret.

Background and light direction converge well for the same reason — a plain light background with a gentle falloff and a key from the upper left are close to measurements.

The shadow and the background tone converge less. Each run samples, so one product's backdrop will be a shade different from another's, and the contact shadow will vary in softness.

Product rendering converges least, because it comes from the source. A range shot on one camera in one session converges tightly; a range assembled from supplier photographs over three years does not, and no instruction reaches that.

When to composite instead

For the strictest kind of consistency, there is a non-generative route and it is worth knowing where the line falls.

Isolate each product with the marketplace or store page and composite it onto a generated backdrop. The backdrop is then identical for every product, the product pixels are exactly the ones photographed, and nothing is invented anywhere.

The cost is that it cannot relight. A product photographed under warm kitchen light composited onto a cool studio sweep still carries the kitchen on its surface, and the mismatch is visible even when the edges are perfect.

So: composite when the sources were shot under similar light and fidelity is paramount. Use this page when the sources disagree and you need them pulled towards each other, accepting that the products are being repainted to get there.

Publication gate

This page ships once the workflow has been run against a range photographed from a fixed camera, a set of individually framed product shots, a range mixing tall and flat shapes, and a range assembled from different sources, with the outputs reviewed side by side and the relative sizes checked against the products' real dimensions.