The first three photos are already spoken for

Sellers tend to judge a generated photo set the way they would judge a contact sheet: lay the shots side by side, keep the ones that look best together, discard the odd one out. That is the wrong test, and it tends to throw away the most useful frame in the set. The shots are not variations competing for one slot. They are several different jobs, and one of them is deliberately built to clash with the others.

Three of the shots are decided before you choose anything

Every set is planned around a mandatory spine: exactly one clean white packshot, exactly one on-model frame, and at least one macro material detail. When the count is too small to fit all three, they are filled in that order — packshot first, then on-model, then macro. Only what is left over goes to the frames most people are actually hoping for: a dramatic hero, a lifestyle shot in context, a feature-proof close-up.

Sets come in three, four or six shots. Read those numbers as a budget for the second group, because the first group takes its cut before anything else is planned:

• Three shots: the spine and nothing else — packshot, on-model, macro. The right choice when you need a listing covered and no more than that.
• Four shots: the spine plus one campaign frame. Decide what that fourth frame is for before you generate, because you get exactly one.
• Six shots: the spine plus three. This is the first count where one set can carry a product page, an ad and an email without reusing a frame.

The white packshot is supposed to look off-brand

The packshot is pure white seamless, no props, no text, long lens — and it is the one frame allowed to break the set's colour grade. Where every other shot holds a shared palette, light direction and grade, the packshot goes high-key, shadowless and neutral on purpose. Next to four moody campaign frames it looks flat and a little cheap, and it is doing its job exactly right.

Its destination is a marketplace listing, which is a different brief from an ad. A plain white background on the primary image is a common marketplace requirement, but the rules differ between them and change without notice, so check the ones you actually sell on rather than trusting any tool's summary of them. What matters here is that the frame is in every set and was never trying to match the campaign. Judging it beside the campaign frames is judging it against the wrong brief, and deleting it for losing that comparison is how a seller ends up with five beautiful photos and nothing to put in the first slot.

The on-model frame exists even when nobody wears your product

On-model is the other guaranteed frame, and it is not only for clothing and jewellery. When the product is not one a person handles — furniture, décor, a large appliance — that shot is staged instead as an art-directed environment with a human present only for scale. The frame therefore answers a different question depending on what you sell. For worn products it answers whether this will fit me. For everything else it answers how big this actually is, which is the question a photograph on its own is worst at.

That is worth knowing before dismissing it. A seller of shelving who reads the person in the frame as a stylistic flourish, and keeps the product-only hero instead, has dropped the one shot in the set that carried scale.

Give each frame a destination before you pick a number

The count is really deciding how many surfaces one set can cover. Write the surfaces down first, then choose:

• Marketplace primary image — the packshot. Nothing else in the set can do this job.
• Product page gallery — the macro, which answers the material question that text cannot: how the leather is grained, how the stone is cut.
• Paid social and display — the hero or the on-model frame, because these have to survive being seen at thumbnail size.
• Email and lifestyle placements — the in-context frame, which is the first thing missing if you ordered three.

Variety comes from the lens, not from walking around the product

One more thing shapes the set, and it is a constraint rather than a choice. The stills are built from the photographs you actually attached, so a face nobody documented is a face the set will not shoot. Given a single photographed viewpoint, the whole set holds that viewpoint and takes its variety from crop, scale, distance, lens, light and staging instead.

Which is why two shots from one angle are not a wasted pair. A 100mm macro and a wide hero from the same viewpoint are further apart, to a shopper, than two three-quarter views would be — and neither of them invents a back panel that was never in front of a camera.

So the sequence that works is: list the surfaces you have to fill, subtract the three the spine already owns, and buy the count that covers what is left. It is a smaller decision than choosing a mood, and it is the one that decides whether you can ship the listing at all.