Use Cases
Category banners, campaign creative and lifestyle imagery — without breaking the catalogue.
Short answer
In e-commerce, use AI around the product rather than for it: category banners, campaign creative, lifestyle scenes and background replacement. The product itself must always be a real photograph — consumer regulation requires the image shown to represent the product sold. Adobe Firefly carries the lowest legal risk for commercial use.
In e-commerce the value of AI image generation is not in product photography but in the visual layer around the product. A shop cannot change its product photos, but it has to produce dozens of campaign banners, category images and social ad creatives a year.
Agency or photo-shoot costs for those images are high and the process is slow. AI creates real economics here: a campaign visual's draft is ready in minutes and the approval loop shortens.
But the legal boundary in this area is clear and must not be crossed: no image showing the product may be generated. The distinction is simple — generate atmosphere, not products.
Decide which images show the product and which set atmosphere. Every image showing the product must be a real photograph; every image setting atmosphere can be AI-generated.
Fix the season campaign's palette and light character as a text block and repeat it across every banner. Visual discontinuity between category pages lowers a site's perceived trustworthiness.
Banners carry a discount figure, a campaign name and a button. Reserve those areas in the prompt; cropping later breaks the composition and cuts the key element on mobile.
Consumer protection law and distance-selling regulation prohibit a seller from giving misleading information about a product. That covers imagery as much as text.
An AI-generated product image is misleading to the extent that it differs from the real product. A jar whose label cannot be read, a texture that does not exist, packaging in different proportions — all of it falls inside that scope.
By contrast, the autumn leaves in a campaign banner, the abstract background in a category image or the living room in a lifestyle scene do not represent the product; they are atmosphere, and generating them is unproblematic.
The practical rule: if a customer forms an expectation about the product from this image, the image must be real. If it only conveys a feeling or a season, it can be generated.
For images that do not represent the product, yes: campaign banners, category images, abstract backgrounds and atmosphere scenes are all fine. Images showing the product itself must be real photographs; consumer regulation prohibits misleading imagery, and AI-generated product images fall inside that.
If you are a corporate brand that cannot carry legal risk, Adobe Firefly — trained on licensed data, it has the lowest infringement risk. If image quality is the priority, Flux gives better results. If the image goes straight into a layout, Canva cuts total time the most.
Desktop banners are wide (around 3:1) while mobile shows something close to square; the same image gets cropped at two ratios. The fix is keeping important elements near the centre, or generating the same scene at two ratios — changing only the ratio parameter in the prompt is enough.
Prompts are written in English because most tools give markedly better results with English prompts.
wide seasonal banner background, autumn palette of rust amber and deep green, soft blurred foliage on the right third, warm low sunlight, clean empty area on the left for headline and button, cinematic but uncluttered, 3:1 ratio
Works well in these tools: adobe-firefly, flux, canva
Note: Wide ratios like 3:1 are right for banners but get cropped on mobile; keep every important element towards the centre.
lifestyle interior scene, a bright living room in late afternoon light, linen sofa and wooden shelves, neutral warm palette, no people, empty coffee table in the foreground for product placement, photographic, shallow depth of field
Works well in these tools: adobe-firefly, flux, canva
Note: Adding "no people" matters: hands and faces are still the details that break most often in generations containing people.
Trained on licensed data; for teams that cannot take legal risk on commercial work.
The strongest open-weight family for photorealistic people and for text rendered inside images.
Image generation, design templates and publish-ready output in one interface, with the generated image landing straight into a layout.
Free credits that renew daily, a large model library and dedicated tooling for game assets.