AI Product Photography

AI product photography means making new photos of a real product without a studio: you start from a photo you already have, and an image model puts the product in a new setting, with new light, props or a lifestyle scene around it. It is for products that exist. For packaging that is still a flat design file, use the AI packaging mockup generator.

There are two ways to do it. The first keeps the product from your photo and replaces only what is around it, the way our AI background generator works, so the angle can’t change. The second asks the model to take the picture again, so the product can stand on a pedestal, lie in a flat lay or sit in a hand. That way changes the composition, and it redraws the product too.

This page does both. A new photo is the default, and every AI product photo opens beside the photo it came from. Keep the product exactly switches to the first way for the shots where the angle can stay.

The model can only keep what it can see, so the photo you upload matters more than the shot you pick. Put the whole product in the frame with a little room around it, sharp, with the front label facing the camera and nothing in front of it. Even light works best: daylight from a window, not a flash that leaves a glare across the label.

Aim for about 1,500 px or more on the long side; the page scales uploads down to 2048 px. A cluttered kitchen behind the product is fine, because the old background is left out of the new picture.

If the label wraps around, add photos of the back and sides after the front. The front photo goes first: it is the side the model turns toward the camera, and the one Keep the product exactly uses.

A new photo is an edit: the model reads your product and draws it again in the new shot, and labels are where that shows. In Photoroom’s benchmark, 850 fashion products went through four image models; the best kept the product accurate in only 29% of its pictures, and logo and text distortion turned up in 20.1% of all 3,400 pictures, the most common fault (Photoroom, July 2026). Packaged goods are not fashion, but their labels are full of text.

So every AI product photo here opens Side by side with your photo. Big words and logos usually come through; check the small print, the net quantity and any claims letter by letter before a buyer sees them.

When every word must stay, switch on Keep the product exactly. Your first photo is cut out (1 more credit), GPT Image 2.5 paints only the scene around it, and our server lays the product’s own pixels back over the picture. The angle stays as you shot it, so In hand and Flat lay, which turn the product, are off in this mode. If the model draws the product somewhere else, that picture is not kept and its credits are returned.

Pick a new photo for a new composition (in a hand, from above, on a pedestal) and keep the product exactly for a crisp label in a new scene. You can use both: new photos for social posts, the exact mode where the label must be read.

Amazon’s product image guide asks for a main image on a pure white background (RGB 255, 255, 255), with the product filling 85% of the image, and allows a lifestyle main image for only a few product types. Other angles and the product in use belong in the other images of the listing.

So split the work. Under any result, Original on white opens the AI background generator with your original photo, for a white main image laid out and checked against that rule. Use AI product photography for the rest: the product in use, in a hand or with its ingredients, plus ads and social posts.

Amazon also asks sellers to tag images with photorealistic people made entirely by AI (the keyword contains-synthetic-performer in the file’s metadata). None of the 8 shots asks for a person; In hand asks for only a hand and wrist, never a face. Whether a hand alone needs the tag is for Amazon to say, so read the guide before you upload an In hand photo there. Amazon reviews every image itself, and we can’t promise a listing will pass.

Studio stands the product on a seamless backdrop in soft light with a reflection beneath it, a clean picture for a product page. Pedestal puts it on a display stand against a pastel backdrop with a plant beside it, an editorial look for a launch.

In use places it where it is used (a kitchen counter, a bathroom vanity or a desk) in window light. In hand holds it up in daylight like a phone photo. Flat lay shoots from straight above among ingredients or contents that suit it. Outdoors sets it on stone, wood or grass in natural light.

Ad banner switches to a wide 16:9 picture with the product on the left third and the rest left calm for a headline. Seasonal adds holiday decor: name the season or holiday in the notes, such as back to school or spring, or you get a winter holiday table.

Any shot takes a line of notes, such as a sprig of mint or late afternoon light. The shot and its instruction stay fixed, so one product in three shots still reads as one set.

Each picture costs credits by model and resolution. GPT Image 2.5 Sunburst and Flare take 10, 12 or 15 at 1K, 1.5K or 2K; Nano Banana 2.1 takes 8 at 1K or 11 at 2K; Seedream 5.0 Pro takes 11 at 1K or 1.5K and 22 at 2K; Muse Image takes 2 at its one size. A run makes 1 to 4 pictures and costs the credits times the pictures, plus 1 for the cutout when you keep the product exactly.

Credits come from a plan or a credit pack; plans start at 800 credits a month (pricing). Every run uses credits, and credits for a run that fails are returned automatically. Photos you add are uploaded and safety checked as you add them, and go to the image model when you press Generate.

Under each result are the next steps. Write the listing opens the AI product description generator with your original photo attached. Edit in chat opens AI Chat with the picture, to change a detail by talking, and Edit puts it back in the composer as a reference.