AI Product Photography: Studio-Quality Shots From a Phone Photo

Picture a candle maker with forty scents and one kitchen table. Every scent needs a cozy lifestyle shot, a clean catalog image and a holiday version. That's a lot of setups for one person with a phone.

AI product photography changes the job. You photograph each product once, upload that photo as a reference, and describe the scene you want. The model keeps your product and builds the room, the surface, the light and the props around it. What it should never do is invent the product itself: a rendered item that doesn't match what ships is a misleading listing, not a photo.

The reference photo is the part you can't fix later, so give it five minutes.

Diffuse light. Near a window on an overcast day, or with a sheet of paper bouncing light onto the shadow side. Hard flash creates glare the model will copy.

Plain background. White, gray or a neutral tabletop, with nothing else in the frame.

Sharp details. Labels, logos and stitching should be readable at full size. If you can't read the label, neither can the model.

Extra angles help. Front, side and back shots can all go in as references. Nano Banana 2.1 accepts up to 14 reference images and GPT Image 2.5 up to 16.

Nano Banana 2.1 is the Image Lab's default and a good first choice for most products: precise edits from references at 2K or 4K for 12 or 18 credits.

GPT Image 2.5 is the pick when packaging is covered in text or the scene needs readable copy. Its Sunburst version is tuned for precise production edits. At 2K, medium quality is 23 credits and high is 56, plus 1 credit per reference image.

Spicy Imagine Pro, powered by Seedream 5.0, is built to keep lighting and color tones from reference images, at 8 credits for 2K or 15 for 4K. It's worth a comparison run on products where the exact color matters, such as cosmetics or textiles.

Run the same reference and prompt on two models and keep the version that reproduces the product most faithfully.

A useful product prompt covers five things: the product (point to your reference), the surface, the setting, the light and the camera, plus one line about what must not change.

Candle: “The candle from the reference image on a reclaimed oak side table, evening living room, a wool throw softly out of focus behind, warm lamp light from the left, 50mm lens at table height. Keep the label, jar and wax color exactly as in the reference.”

Sneaker: “The sneaker from the reference on wet concrete at dusk, city lights blurred in the background, low camera angle, cool blue ambient light with a warm rim light from behind. Keep the colorway, logo and laces unchanged.”

Serum bottle: “The bottle from the reference on a pale stone bathroom shelf, morning daylight through frosted glass, a sprig of eucalyptus beside it, overhead three-quarter angle, soft shadows. Keep the label text exactly as in the reference.”

Match the light. If your reference was shot in cool daylight, a warm sunset scene will make the product look pasted in. Describe light that could plausibly have lit your original photo.

Write a scene recipe. For each product category, note the surface, palette, light and camera angle that worked, and reuse it word for word. Changing only the product keeps a catalog coherent.

Use presets for speed. Product templates give you finished looks at 18 credits per image at 1K or 2K, and Multi-Shot turns one reference into several cinematic variations at 16 credits per shot.

Make the plain version too. Marketplace main images usually need a clean background. Background Remover cuts out a product for 4 credits.

Prepare for zoom. Image Upscale takes your keepers to 2K, 4K or 8K for 8, 10 or 12 credits.

Compare against the real item. Color, proportions, label text, number of pieces, included accessories. Models sometimes tidy up small text or add a lid that doesn't exist.

Follow marketplace rules. Amazon, Etsy and Shopify each treat AI images differently; our guide to Amazon image requirements and AI product photos covers them as of October 2026.

Stay on the right side of consumer law. The FTC in the US, Canada's Competition Bureau and Australia's ACCC all prohibit misleading product representations, whoever made the image.

For template-based workflows, see AI templates for e-commerce, and when you want motion, product videos from one photo.

A realistic budget per product: two Nano Banana 2.1 tries at 2K (24 credits) and a 4K upscale of the keeper (10 credits), so 34 credits. Twenty products come to 680 credits, which fits Spicy Pro (1,800 credits for €25 a month) or Spicy Basic (600 credits for €10) plus a top-up. New accounts get 25 free credits to test the workflow on one product; see the pricing page for plans.