AI Product Photo Generator

Why references beat descriptions

Ask a model for “a black energy drink can with a purple logo” and you get a can — a different one every time. Give it a photo of your can and you get yours. That is the whole trick of AI product photography: references in, product preserved. Nano Banana Pro and GPT Image 2 each accept up to four reference images, so you can show the front, the back and a close-up of the label.

The sentence that protects your packaging

Start every product prompt with “keep the product, logo and lettering unchanged”. Then describe only what should change: the surface, the props, the light, the background, the camera angle. The before-and-after on this page used a cluttered, flash-lit kitchen-counter snapshot as the only reference; the result kept the can and its lettering and replaced everything around it.

Light is the luxury

What makes a product photo look expensive is mostly lighting. Name it: “soft window light from the left”, “hard top light with a long shadow”, “violet rim light”, “golden hour backlight”. Add one texture the eye can feel — wet slate, linen, travertine, brushed metal — and one or two props at most. Clutter reads as cheap.

Sizes and next steps

Nano Banana Pro renders 1K, 2K or 4K, in ratios from 1:1 listings to 4:5 feed ads, 9:16 stories and 21:9 banners. When a still is right, it is also the perfect start frame for image-to-video: a slow orbit, condensation running down, a splash bursting into motion. That is how the product clips on this page were made.

Use photos of products you own or sell, and check marketplace rules on generated imagery before you list.

Build a consistent set

Store listings and ads look professional when the photos match each other. Once a scene works, keep its wording and only change the product or the angle, so the light, surface and colour grade carry across the whole range — one prompt, one look, every SKU.