How Agencies Use AI to Produce Product Photos for Multiple Clients (2026 Guide)

An agency's product photography problem is not one product. It is one product for each of many clients, each with its own look, its own approver, its own channels and its own idea of what "on brand" means — and a rule that nothing from one client's work may ever appear in another's. Generating images does not solve that; it makes it easier to get wrong faster. What solves it is an operation: a library per client, a look per client, one check for every image, and a record that says what each delivered file was made from.

This guide is that operation. It publishes no figure for what agencies save or charge; the comparison of what shoots, freelancers and tools cost is in product photography pricing, and packaging the work as a service is in how studios can add AI product photography services. This page is about running many clients through one pipeline without mixing them up.

How do agencies produce product photos for multiple clients with AI?

By keeping clients separate at the level of the library, not the file name. Each client gets its own workspace, holding its masters — the client's products photographed on a plain background with even light and a colour reference — its approved wearers where there are any, and one saved look: scene, surface, light, props rule and a brand reference image. Every image for that client is generated from that client's masters into that client's look, compared to its master at full size before it is shown to anyone, proofed to the client with a note of which files are derived, and delivered with a record. The next client is a different workspace with a different look; the pipeline is the same.

What is multi-client AI product photography?

It is the production of product images for several brands from one pipeline, where each brand's sources, look, approvals and outputs are kept apart and every delivered image can be traced to the photograph it came from. It differs from a single brand's workflow in the number of things that must not leak — masters, wearers, looks, references — and from a studio's work in that the agency usually did not take the photograph and has to make sure a usable one exists before anything is generated.

One workspace per client

• Masters. The client's products, photographed to be derived from: plain background, even light, a colour reference in frame, every side the catalogue needs, full resolution. If the client supplies photographs that do not meet this, the first deliverable is a request for ones that do, or a master shoot; generating from a poor source produces a poor image for every channel.
• Wearers. For apparel and jewellery, the model the client approved — from the catalogue or one the client photographs and keeps — and only that model across the client's range.
• The look. Scene, surface, light, props rule and a brand reference image, saved once and reused for every product, so the client's catalogue reads as one catalogue.
• The library. The client's approved images in folders the client can understand — by range, by season, by channel — and nothing from any other client in it.
• The people. The agency members who work on that client, invited into that client's workspace and no other.

In Photta a workspace is the unit: its own library and folders, its own members, its own saved wearers and looks, and credits that belong to it. Switching client is switching workspace, which is the point.

The pipeline, per client

1. Brief. Which products, for which channels, in which slots — main image, gallery, scene, scale, ad — and which look. A brief that does not name the master is sent back.
2. Source. The master exists and passes: background, light, colour reference, sides, resolution. If not, stop here.
3. Generate. Into the client's saved look, one change at a time where the product's fidelity matters, a few variants where the choice is creative.
4. Check. Every image against its master at full size — outline, surface, label, colour against the reference, one contact shadow agreeing with the scene — by someone who did not generate it.
5. Proof. The set to the client with the record: which files are derived from which photograph, which wearer, which look.
6. Deliver. Approved images cropped per channel, named per product and channel, with the record kept in the client's workspace.
7. Reorder. Next season is the same masters in a new look; new products are new masters through the same look.

Multi-brand catalogues: one recipe per brand, one pipeline

A retailer or a group that sells several brands has the agency's problem from the inside: many looks, one production line, and a page where products from different brands sit side by side and must not look like they came from different planets. The structure is the same as for many clients — a workspace per brand — with two additions.

First, a house layer above the brand recipes. Each brand keeps its own scene family, palette and props, but the listing slots a customer compares across brands — the main image on white, the grid tile, the crop — follow one house rule, so the category page reads as one store even where the lifestyle images read as several brands.

Second, a single review standard. Every brand's images are compared to their masters by the same rows, with the same classes of rejection, whoever produced them, so that quality is a property of the pipeline and not of whichever brand's team was more careful that month.

What stays separate: masters, wearers, scene descriptions, credits and records, per brand, so a brand can be added, paused or handed back without touching the others. What is shared: the house slots, the review rows and the export naming, so the catalogue behaves as one catalogue.

What the client sees, and what they never see

• Before generation: the master requirements, the wearer for approval, the look as a reference set.
• At proof: the derived images beside their masters, and a plain sentence on what is a photograph and what is derived.
• At delivery: the files, the record, and what is excluded — a change to the product itself is a reshoot, not a generation.
• Never: another client's products, wearers, looks or references in their proof, their library or their images. A prop that happens to be another client's product is a failed check.

Workspace, look, record

The loop the checklist below serves, named so the other agency pages can point at it.

1. Workspace — one per client, holding its masters, wearers, look, library and members, and nothing of anyone else's.
2. Look — the client's scene, surface, light, props rule and reference, defined once and reused across the range.
3. Record — every delivered image traced to its master, wearer, look and date, kept where the client's images live.

Multi-client production checklist

The separation row is the one an agency cannot recover from. Everything else is regenerated; a leak is not.

Two clients, one afternoon

An agency handles a skincare brand and a ceramics brand. Each has a workspace. The skincare brand's masters were shot by the brand on light grey with a grey card; its look is pale stone in soft side light with a folded cloth. The ceramics brand's masters came in as busy kitchen photographs, so the agency's first deliverable was a master shoot on white; its look is reclaimed wood in morning light with a single plant.

For each brand the producer briefs the slots, the generator runs the range into the saved look, and a second person compares every image to its master: the serum's label colour against the grey card, the mug's glaze against the white. The skincare proof goes to its client with the record; the ceramics proof to its client, from its own workspace, with its own record. Neither client's proof contains a prop, a surface or a colour from the other's look. Next season each brand gets a new look on the same masters, and the agency's afternoon looks the same.

Creative agencies: campaign work beside catalogue work

A creative agency's clients want two things that look alike and are produced differently: the catalogue set, where the product is the subject and nothing may change, and the campaign set, where a scene, a mood and a message are the subject and the product is a kept guest.

Run them as two lanes in the same workspace. The catalogue lane is the pipeline this page describes: masters, recipe, review by class, exports per channel. The campaign lane starts from the same masters and adds a written brief per campaign — placements, formats, palette, copy space — generates each format at its own framing around the kept product, and makes concept visuals that carry no product at all from prompts and reference photographs. The line between the lanes is the product's pixels: a campaign scene may be anything the brief wants, and the product inside it may not be anything but itself.

What the agency prices differently is the brief and the review: a campaign set is judged against the brief and the master, a catalogue set against the recipe and the master. What it never does in either lane is generate the product from a description, because the client's customer compares the image to the box, and a persuasive scene invites a persuasive product.

Frequently asked questions

How do agencies keep client product photos separate when using AI tools?

By keeping each client in its own workspace: its own masters, approved wearers, saved look, library, members and credits. Separation at the file-name level fails as soon as two people are working; separation at the workspace level is structural. A prop, wearer or reference from one client appearing in another's image is a failed check, not a style choice.

What does an agency need from a client before generating product images?

Masters: the client's products photographed on a plain background with even light and a colour reference in frame, every side the catalogue needs, at full resolution. If the client's photographs do not meet that, the first deliverable is a request for ones that do, or a master shoot. Also an approved wearer where the range needs one, and the brand's look as a reference set.

How do agencies keep AI product images on brand for each client?

With a look saved once per client — scene, surface, light, props rule and a brand reference image — and reused for every product, rather than a scene described per image. On-brand across a range is a property of the saved look, not of individual prompts.

Should the client be told which images are AI-derived?

Yes, at proof and at delivery, with a record per file: the photograph it came from, the wearer, the look and the date. Marketplaces and the client's own customers will ask; the agency that can answer file by file keeps the client.

How should an agency price AI product photography for clients?

By the deliverable and the checking attached to it — a named set of images compared to the master by someone who did not generate them, delivered with a record — rather than by generation cost. The master shoot stays priced as photography. This page gives no figures; the packaging of the offer is covered in the studio services guide linked from this page.

What this page does not claim

That AI production raises an agency's margin by any amount, or that clients prefer derived images; this page measures neither. What it claims is that the multi-client problem is separation and record-keeping rather than generation, that a workspace per client is the unit that makes separation structural, and that every delivered image must be traceable to a photograph.

To run a client's range from its masters, see AI product photography; for the offer and its pricing logic, how studios can add AI product photography services; for the roles that run the pipeline at scale, building an AI creative operations team.

Companion guides

• AI Product Photography for Photography Studios
• Building an AI Creative Operations Team