What counts as a set?

Exactly what the quote says it is, because this is the single most common source of disappointment in the category. A set is defined by the number of products, the number of angles per product, and whether lifestyle scenes are included as well as plain-background listing shots. We write that out in full so nobody is imagining thirty images while somebody else is producing eight.

Should we use the camera route or the AI route?

It depends on what the images have to prove. Anything where fine texture, exact finish or physical detail is the selling point benefits from being photographed. Anything where the product is already documented and what you need is volume — the same item in many settings, seasonal backgrounds, banner crops — is faster and cheaper generated. Many catalogues are best served by both.

Will AI images pass marketplace listing rules?

Marketplace rules are about the image, not the method — pure white backgrounds where required, the product filling the frame, no added text or badges on the main image, correct proportions. Generated images can meet all of that, but they have to be checked against it deliberately. We work to those constraints rather than producing something attractive that then gets rejected.

How do you stop AI mangling our packaging?

By checking it, every time, at full zoom. Image models routinely corrupt small text — a brand name loses a letter, an ingredients list turns into plausible nonsense, a logo shifts shape slightly. Every generated image is inspected at full zoom across every word and mark, and where the label is wrong the real label is composited back in from your own photograph.

Will you tell us which images are AI generated?

Always, in writing, as part of the delivery. It matters for your own compliance decisions, for marketplaces that ask, and for advertising rules that are tightening in several markets around synthetic imagery — particularly where a human figure appears in a lifestyle shot. Knowing which file is which protects you, so it is documented rather than assumed.

We already have photos taken on a phone. Is that enough to work from?

Often yes, for the AI route. Clear, sharp, straight-on shots of the product from a few angles, with the packaging legible, give enough to work from — background removed, then placed into properly lit scenes. What they cannot be is blurry, heavily filtered or shot at an extreme angle, because errors in the source propagate into everything generated from it.

Video, Audio & Media Production

Product photography, camera and AI

Listing images, detail shots and lifestyle scenes produced at catalogue scale — shot in studio, generated with AI, or both, with every frame checked against the real product.

From: $100–$3,000/set
Typical time: 1–3 weeks
Usually bought by: E-commerce

What this actually solves

An online buyer cannot pick the product up. The photograph is not a representation of the product, it is the product as far as the purchase decision goes — and everything the shopper cannot verify by touch has to be carried by the image.

The person buying this rarely describes it that way. What they describe is a blockage. New lines are sitting in the warehouse and cannot be listed because there are no images. A marketplace has rejected a listing for a background that is not white enough. A seasonal campaign needs the same products against a different setting and there is no time to reshoot.

Underneath most of these is the real problem, which is scale. A large catalogue with a steady stream of new arrivals every month does not have an image problem, it has a throughput problem. What they need is not a photograph, it is a repeatable process that produces consistent images every time new stock arrives — same lighting, same crop, same background, same naming, so the site looks like one shop rather than a collection of uploads.

That is why there are two routes here and why the choice between them is a practical one rather than an ideological one.

How the work runs

Define the set and build the shot matrix. Which products, how many angles each, which are plain-background listing images and which are lifestyle, what the marketplace rules require, and how files should be named so they map to your SKUs on import. Half a day.
Capture or source. On the camera route — a lightbox or a white sweep so there is no visible corner behind the product, two lights, a tripod, and four to eight angles per item, with focus stacking on anything small where a single frame cannot hold the whole object sharp. On the AI route — your existing product photographs prepared, backgrounds removed, and scenes generated. One to five days.
The base pass, applied across everything. Exposure, white balance against a known reference, background brought to a true clean white where required, dust and scuffs removed. Done as a batch so the set is internally consistent before anyone starts on individual images. One to four days.
Per-image work. Focus stacks blended, generated scenes colour-matched to the rest of the set, and any corrupted packaging or label rebuilt by compositing the real artwork back over the generated version. Two to seven days.
Full-zoom quality check against the real product. Every image inspected at full zoom, every word on every label read, every logo compared to the original. It is a small amount of attention per image and it is the step that separates professional output from embarrassing output. Two to four hours.
Export, name and deliver. Web sizes at around 2000px on the long edge plus full-resolution masters for print, named to your convention, with a written note of which images were generated. Half a day.

Worked in Lightroom or Affinity Photo, with focus stacks blended automatically and then corrected by hand, and generated scenes produced through current image models and cleaned up in a proper editor rather than shipped as they come out.

What moves the price

Sets are quoted at $100–$3,000/set and typically run 1–3 weeks, with three factors deciding where inside that a particular catalogue lands.

How many products and how many angles each. This is the main driver and it is close to linear. The count of products multiplied by the count of angles gives the count of images, and images are what take the time.

Whether lifestyle scenes are included. A plain-background listing image is a solved, repeatable problem. A product in a convincing environment — the right surface, the right light direction, the right props, consistent across a whole range — is a creative job each time, whether it is built in a studio or generated.

The condition of what you supply. Good reference photographs, a vector logo and the actual packaging artwork make the work fast and accurate. Low-resolution images pulled from an old website mean rebuilding before producing.

What usually goes wrong

Generated images ship without anyone reading the words on them. This is the defining failure of the AI route. The model produces a beautiful bottle and renders the brand name with a letter changed, or turns the ingredients panel into convincing gibberish, or subtly redraws the logo. A shopper might not notice. The brand owner notices instantly, and it destroys confidence in the whole batch. Every generated frame gets read at full zoom, and where the text is wrong the real label goes back on.

Nobody defines what a set contains. You picture a complete product page worth of images and the quote covered a handful. Neither party is being dishonest — the word was never pinned down. The count of products, the count of angles, and whether lifestyle is included go into the quote as numbers.

Colour is treated as a look instead of as a fact. A warm, fashionable grade makes the images more appealing and turns your navy into teal. Your customer receives something that does not match what they saw, returns it, and the return is blamed on the product. Product work is the one place where colour fidelity outranks style, which is why a white balance reference is shot or set at the start and the whole set is matched to the physical item rather than to a mood.

See where the site stands before you brief anyone.

If the site is already live, start from what it scores rather than from a guess. The audit reports what it checked, what it found, and — the part most tools leave out — what it could not check, and why.

See the website audit Talk about this work

Questions people ask before they commit.

What counts as a set?: Exactly what the quote says it is, because this is the single most common source of disappointment in the category. A set is defined by the number of products, the number of angles per product, and whether lifestyle scenes are included as well as plain-background listing shots. We write that out in full so nobody is imagining thirty images while somebody else is producing eight.
Should we use the camera route or the AI route?: It depends on what the images have to prove. Anything where fine texture, exact finish or physical detail is the selling point benefits from being photographed. Anything where the product is already documented and what you need is volume — the same item in many settings, seasonal backgrounds, banner crops — is faster and cheaper generated. Many catalogues are best served by both.
Will AI images pass marketplace listing rules?: Marketplace rules are about the image, not the method — pure white backgrounds where required, the product filling the frame, no added text or badges on the main image, correct proportions. Generated images can meet all of that, but they have to be checked against it deliberately. We work to those constraints rather than producing something attractive that then gets rejected.
How do you stop AI mangling our packaging?: By checking it, every time, at full zoom. Image models routinely corrupt small text — a brand name loses a letter, an ingredients list turns into plausible nonsense, a logo shifts shape slightly. Every generated image is inspected at full zoom across every word and mark, and where the label is wrong the real label is composited back in from your own photograph.
Will you tell us which images are AI generated?: Always, in writing, as part of the delivery. It matters for your own compliance decisions, for marketplaces that ask, and for advertising rules that are tightening in several markets around synthetic imagery — particularly where a human figure appears in a lifestyle shot. Knowing which file is which protects you, so it is documented rather than assumed.
We already have photos taken on a phone. Is that enough to work from?: Often yes, for the AI route. Clear, sharp, straight-on shots of the product from a few angles, with the packaging legible, give enough to work from — background removed, then placed into properly lit scenes. What they cannot be is blurry, heavily filtered or shot at an extreme angle, because errors in the source propagate into everything generated from it.

Often taken together.

Start with the brief, not the invoice.

Send what you are trying to achieve. You get a written scope, a price and a date — or an honest “this is not the right service for that”.

Talk about Product photography (including AI) All Video, Audio & Media Production services