INSIGHTS

How do you use AI for product photos without changing the product?

Start with one approved packshot, then brief and check the listing image, paid social crop and lifestyle scene separately.

Listen to this articleAI-generated voice
A designer edits the background of a mint tea product photo in Photoshop, with the carton beside the monitor.

By Catarina Mestre

·

5 min

·

11 November 2026

Keep the approved product photograph on its own layer and use AI to make the scene around it. Compare each finished crop with the original before it goes to the client. Check the label, colour and shape, and make sure the scene shows what the buyer will actually receive.

A single product photo can serve a listing, a paid social ad and a lifestyle scene. Here is how to brief those three files and decide which are ready to send.

What should you lock before making the first image?

Start with an approved packshot, a clear photograph of the item as it will be sold. Put the original in a shared folder with the product name, variant, file version and approved label artwork. Check the brief for the client's rules on AI use before sharing their files with a tool.

Take a fictional job for a tea brand. The client has approved a front-facing photograph of a pale yellow mint tea carton with a dark green panel. The label reads “MINT TEA” and “20 BAGS”. The brief allows a new background but keeps the carton shape, print, colour and quantity fixed.

The photograph establishes what the front of this carton looks like. If the client asks for a shot that shows the side, request a matching photograph of that side. A convincing guess can put the wrong text on the box.

Which parts of a product photo can AI change?

Generate the background on a separate layer beneath the cut-out product photograph. Keep the original and name the layer so the next designer can find it. That gives the team a clean source to return to when a crop or shadow needs work.

Adobe's Photoshop background guide documents a way to make a background around a selected subject. Google's Product Studio guide also describes scene generation and warns that outputs can be unexpected. Tool features checked on 11 November 2026.

Brief the scene in terms the designer can check. For the tea carton, ask for a light counter and a soft shadow in the same direction as the light on the carton. A cup of brewed tea can suggest use. A second carton can suggest that the price buys two.

How should you brief the three versions?

Give each file one job. The shop listing helps the buyer identify the exact carton. The ad needs a crop that still shows what is being sold on a phone. The lifestyle image shows where the product fits into someone's day.

For Google Merchant Center, check the rules for the destination field. Google's main image requirements require the exact item and variant. Its additional image requirements allow a wider range of views, including the product in use. Both sets of rules were checked on 11 November 2026.

Give the client a contact sheet with the original on the left and the three final exports alongside it. Include the placement and crop under each one. This is the approval sheet the producer keeps with the job.

One carton, three briefs

Worked example for a fictional mint tea brand.

PlacementWhat may changeWhat has to matchWho signs off
Main listing imageClean crop and spacingOne yellow carton, green panel, exact label and quantityListing owner
Paid social cropFraming, layout and backgroundSame carton and the offer shown in the adArt director and campaign lead
Lifestyle imageCounter, lighting and a cup of teaSame carton, credible size and clear contents of the purchaseArt director and client

How do you catch a changed product before sign-off?

Compare the final exports side by side with the approved photograph, first at the same size and then at the size people will see. At the larger size, read the label and check the shape, colour and finish. At phone size, check that the crop still makes the product and the offer clear.

Suppose the social crop hides “20 BAGS” just as the ad copy introduces a multi-buy offer. Change the crop or clarify the offer before approving that file. If the lifestyle scene adds a second carton in the background, remove it or use a different scene. Keep the approved listing photograph available while the other two are corrected.

What to do when an export fails a check

Record the decision on the file the team will actually deliver.

What changedWhat to do
Label or shapeRestore the approved product layer and export again.
Crop or offerReframe the asset and check it with the copy at phone size.
Props or quantityRemove anything that could be mistaken for part of the purchase.
Colour or missing detailCompare with the approved source. Request another photograph if it cannot settle the question.

What should the producer send for approval?

Send the approved source, the final files and the placement notes together. Give the art director responsibility for matching the product. Ask the person who owns the listing or campaign to check what the viewer will see on that placement.

Check the files again after resizing and exporting. Google's guidance on AI-generated content requires embedded information that identifies AI-generated imagery. It also says to preserve existing tags that describe how an image was made. Confirm that the final file still carries the relevant information if you are supplying it to Merchant Center. Guidance checked on 11 November 2026.

If the product itself has changed, return to the approved photograph and rebuild the scene around it. Ask for a new photograph if the team cannot establish what an unseen part looks like. Use the same review process for other client work when someone needs to make the final call.

What should you do on the next brief?

Pick one product and ask for the approved photo before anyone generates a scene. Put the three placements and their approval owners in the brief. For each one, write the product details that must match and the changes the client has approved.

The Image job in ViraOps Create can help turn those decisions into a brief your designer can use. Send that brief with the approved photograph. Both should travel with the job through sign-off.

Give your designer a brief with the approved product details and the changes each placement allows. The Image job in ViraOps Create helps you write it.

PUT IT INTO PRACTICE

ViraOps turns this into weekly steps for your whole team.

Questions

An AI-assisted product image can keep the packaging, label and shade accurate when the final file uses the approved product photograph. Generate the scene on a separate layer. Compare the exported image with the original at the same size so you can catch changes introduced during editing and export.

Google Merchant Center accepts images made with generative AI when they meet the rules for the image field and carry the required source information. The main image must show the exact item and variant. Check the file you plan to deliver, since an export can affect both the crop and the information embedded in it.

One product photo can serve several crops and backgrounds at its original angle. Use another approved photograph when a new angle reveals part of the product the source does not show. For a carton, that might mean getting the side panel before you show its text or construction.

SOURCES