How to Improve Product Photography with Nano Banana Pro and AI

Nano Banana Pro is the name Google uses for its Gemini 3 Pro Image model, announced on November 20, 2025. It generates and edits images from text and reference photos, and it is available in the Gemini app, Google AI Studio and the Gemini API, Google Ads, and Workspace apps such as Slides. For product photography, it is one of the most capable general-purpose image models you can use today.

"General-purpose" is the important word. It can do almost anything, which means the quality of your product photos depends on how you set up the job. This guide covers a workflow that works, where it breaks, and when a dedicated product-photo app is the simpler route.

What Nano Banana Pro Can Do for Product Shots

According to Google's announcement and the Gemini API image generation docs, the capabilities that matter most for sellers are:

• Reference images. Up to 14 reference images per request; the API docs say up to 10 of those can be objects kept at high fidelity. In practice, that means you can give it several angles of the same product.
• Resolution. 1K, 2K, and 4K output.
• Aspect ratios. Including 1:1, 4:5, 3:4, 2:3, 16:9, and 9:16, which covers marketplace squares, Instagram portrait posts, and vertical stories.
• Text rendering. Google highlights legible text in images, which helps with packaging mockups and ad creatives.
• Local edits. Changing lighting, camera angle, focus, and color grading on an existing image.

Step 1: Shoot Reference Photos Worth Copying

The model can only preserve what it can see. Supplier photos are usually low resolution, compressed, and already edited, so start with your own:

1. Use soft, indirect window light from one side. Avoid direct sun, which creates hard shadows the model may try to keep.
2. Shoot three angles: front, three-quarter, and back or side. Include a close-up of the label or logo.
3. Keep the background plain (white paper or a wall) so the product outline is unambiguous.
4. Hold the phone at product height and step back slightly, using the 2x lens if you have one. Wide-angle close-ups distort proportions.
5. Check focus on the text. If you can't read the label in your photo, the model can't either, and it may invent something that looks plausible.

Step 2: Write Prompts Like a Shot Brief

Vague prompts ("make this look professional") produce generic images. Describe the shot the way you would brief a photographer: subject, surface, background, light, camera, and what must not change.

White-background listing shot:

"Use the attached photos of this bottle. Place it upright, centered, on a seamless pure white background. Soft, even studio lighting from the front left, a faint natural contact shadow under the base. Straight-on camera at product height. Keep the label text, logo, cap color, and bottle shape exactly as in the reference. No props, no text. Square format."

Lifestyle scene:

"Place this exact bottle on a light oak bathroom shelf next to a folded white towel, morning window light from the right, shallow depth of field with the background softly blurred. 4:5 portrait. Do not change the label, proportions, or colors of the product."

Three habits help: name the details that must not change, specify the light direction so shadows are consistent, and state the aspect ratio for the place the image will be used.

Step 3: Check Every Output Like a Buyer Would

General models are good at making images look right at a glance. Your job is to catch what is wrong at 100% zoom:

• Label text. Read every word. Small print, ingredient lists, and model numbers are where errors appear.
• Shape and proportions. Compare the silhouette to your reference. Bottles get slimmer, boxes lose a bevel, handles move.
• Color. Compare against the real product in daylight, not against your reference photo.
• Reflections. Glass and metal should reflect the new scene, not your kitchen.
• Extra parts. Generated scenes sometimes add a second product, a lid, or an accessory that doesn't ship with the item.

If the result isn't accurate, don't use it for a listing. An image that shows a slightly different product than the one you ship leads to returns and complaints.

Step 4: Handle Watermarks and Metadata

Google's announcement says all images generated with its tools carry an invisible SynthID watermark. It also says images made in the Gemini app by free and Google AI Pro tier users keep a visible Gemini watermark, which is removed for Ultra subscribers and in Google AI Studio. Plan for this:

• Don't upload visibly watermarked images to marketplaces. Google Merchant Center's image requirements prohibit watermarks, and Amazon's image guide limits what can appear on product images. Generate in Google AI Studio, where the visible mark is removed, if you need clean files.
• Keep AI metadata intact. Merchant Center asks that generative AI images carry metadata such as the IPTC DigitalSourceType property, and says not to remove it.
• Disclose AI-generated people on Amazon. Amazon's product image guide requires the contains-synthetic-performer metadata keyword on images with photorealistic, fully AI-generated people.

Common Mistakes

1. Using one reference photo. With a single angle, the model has to guess the rest of the product. Give it at least two.
2. Asking for everything at once. "Change the background, the lighting, the angle, and add props" in one step drifts further from your product. Change one or two things per edit.
3. Mixing light directions. If your reference is lit from the left and the prompt asks for sunset from the right, shadows on the product won't match the scene.
4. Generating at a small size and upscaling. Choose 2K or 4K up front if the image is for zoomable listings.
5. Skipping the real-product comparison. Every image should be checked against the item on your desk.

Where Studio Zero Fits

Writing shot briefs, managing reference images, and checking output is a real workflow, and some sellers enjoy that control. Others would rather not write prompts at all. Studio Zero is a phone app for that second group: you photograph the product and choose one of three modes, Studio for clean white or plain backgrounds, Concept for lifestyle scenes, or Fashion for clothing and accessories including on-model shots. It is free to try on iOS and Android.

It is not a replacement for everything Nano Banana Pro does. If you need text-heavy ad creatives, infographics, or packaging mockups with precise wording, prompting Nano Banana Pro directly gives you more control.

When Neither Is the Right Tool

• Food delivery menus. DoorDash's merchant photo guidelines reject photoshopped or artificial backgrounds for menu items. Shoot the real dish.
• Close-ups that sell on texture. Fabric weave, wood grain, gemstone clarity: shoot these for real.
• Products you haven't received. Generating images from supplier photos without handling the product means you can't verify color, size, or finish. Order a sample first.