GuidesMake the creative

How to do AI product photography for ads

Start with a sharp photo of your real product on a plain background, then use an AI tool to change the scene around it: new backgrounds, lifestyle settings, seasonal versions. Keep the product itself unchanged, inspect every output for shape, color and text errors, and never show customers something you do not sell.

Updated November 16, 20265 steps5 min read

Write my next ads →

$1 today. Cancel anytime before day 7.

From one reference photo to a set of ads
  1. 1Shoot a clean referenceEven light, plain background, product fully in frame.
  2. 2Decide each image’s jobHero, lifestyle, detail, scale, or variant.
  3. 3Generate scenesChange the background and setting, not the product.
  4. 4Inspect the outputCheck edges, color, text, shadows and proportions.
  5. 5Export per placementCrop to the sizes each ad format needs.
On this page
  1. Shoot a reference photo the AI can trust
  2. Decide what each image has to do
  3. Generate scenes, but keep the product fixed
  4. Inspect every image before it runs
  5. Turn the images into ads and test them
  6. Questions

Shoot a reference photo the AI can trust

The quality of AI product photography is set before you open any tool. A phone photo is enough if it has even light, sharp focus, no clutter, and the whole product in frame. Shoot a front view, a three-quarter view and a close detail, and keep the camera level.

A plain white or neutral background makes it easy for tools to cut the product out cleanly. Keep any labels facing the camera and legible, because small text is what AI most often gets wrong when it has to redraw it.

If the product is shiny or transparent, such as glass, jewelry or bottles, light it from the side with a large soft source and use a white card to control reflections. Matte products are far easier. Take a few extra angles even if you only plan one scene, because some tools use multiple views to keep the product consistent. Clean the product first; dust and fingerprints are tedious to remove afterward.

Decide what each image has to do

Different placements want different images. A hero image on a clean background for shopping listings and product pages. A lifestyle image with the product in use for social ads. A detail shot for texture or material, and a scale shot so people know the size. Write the list first, then generate to the list.

Marketplaces and shopping feeds have their own image rules, such as clean backgrounds and no overlays on the main image. If you run Shopping campaigns, read the product image requirements before replacing a hero shot with a generated scene; see Google Shopping campaigns.

A useful set for a small store is five images per product: hero on white, one in use, one detail, one scale reference, and one showing the pack or what is in the box. Generate the lifestyle image for ads and social, and keep the hero a real photo or an unaltered cutout. That split lowers the chance of a policy flag on a feed and keeps the listing true to what ships.

Need the next round of ads?

Pulls competitors’ live ads and breaks down their hooks. Watches your ads for creative fatigue every week.

Generate scenes, but keep the product fixed

Give the tool the reference photo and describe the setting in physical terms: “on a oak table beside a window, morning light, shallow depth of field.” Ask for several versions and for different aspect ratios. Free tiers exist for several tools and are fine for tests, but watch for watermarks, resolution caps and usage terms that restrict commercial use.

If a tool redraws the product rather than placing it, results drift: logos change, proportions shift, a zipper moves. Prefer modes that keep the product pixels and generate only the surroundings, and run a background-change as a first pass before trying full scene generation.

Be specific about physical facts: surface material, light direction, color temperature, and how large the product appears in the frame. Ask for the camera height and a shallow or deep depth of field if the tool allows it. Generate eight to ten options and discard most of them. The hit rate on AI scenes is low per image, but the cost per image is small, so volume plus a ruthless eye is the method.

Inspect every image before it runs

Zoom in to full size and check edges, shadows, reflections, color accuracy and any text. Compare the color with the real product under normal light. An ad whose product looks different on arrival produces returns and bad reviews, which cost more than a photographer.

Also check what the scene implies. A generated lifestyle photo with accessories, props or a result that you do not provide can read as a promise. Remove it. The wider process for generated creative is in AI generated ads.

Compare against a real photo side by side at the same zoom. Typical flaws are a logo that changes letters, a strap or seam that moves, handles that bend and shadows that fall in two directions. If a fix needs more than a quick retouch, regenerate instead. Keep the original reference file with the finished set so you can show what the product really looks like if anyone asks.

Turn the images into ads and test them

Place the best images into static ads and carousels, one angle per image, and test them against your current photography. For ecommerce accounts, Facebook ads for ecommerce covers the catalog and creative setup around them.

Sprites can read your store and ad results, draft copy to go with each image, and propose the changes as approval cards you edit. The images themselves you should review, since you know what the product really looks like.

Label the files clearly with the product, the scene and the date, and note which tool and settings produced each. When an image wins in an ad, you will want to make more like it. Rotate scenes seasonally and by audience: the same product on a kitchen counter for one audience and on a desk for another can be a clean way to test who the product is for.

Questions

Can I do AI product photography for free?

Yes, several tools have free tiers that cover background changes and basic scenes. Check for watermarks, resolution limits and commercial-use terms before using the output in ads.

Are AI product photos allowed in Google Shopping?

Google’s product image rules apply to every image regardless of how it was made, including requirements on the main image. Check the current requirements and keep the main image an accurate, clean shot of the product.

Do I still need a real photo?

Yes. The most accurate AI product images start from a real photo of the actual product. Generating a product from a text prompt alone invents details you do not sell.

What makes AI product images look fake?

Inconsistent lighting, impossible shadows, warped text and surfaces that are too smooth. Look at shadows and reflections first.

What prompts work for product photography?

Describe a physical scene: surface, light direction, time of day, props and camera distance. Avoid adjectives like “stunning” and name concrete things instead.

Keep going

Have Sprites write your next ads

  • Pulls competitors’ live ads and breaks down their hooks
  • Writes ad copy and concepts for your offer
  • Watches your ads for creative fatigue every week