Nano Banana Pro Product Photography: The $0.11 Studio

Nano Banana Pro product photography is the practice of generating studio-grade product images with Google's gemini-3-pro-image model instead of booking a physical shoot. You describe the product, the surface it sits on, the light, and the camera angle in plain sentences, and the model returns a packshot, a lifestyle scene, or a macro detail shot in under a minute. On BananaBanana that costs $0.11 per image.

The quick answer, if you only need one thing from this page: copy the template from Google's image generation docs and fill in the brackets. "A high-resolution, studio-lit product photograph of [product] on [surface]. The lighting is a [lighting setup]. [Camera angle]. Ultra-realistic, with sharp focus on [key detail]." That single sentence structure produced every product shot in this article, first attempt each time.

To keep the demos honest, everything below is built around one imaginary product: a minimalist wristwatch with a brushed steel case, a forest-green sunburst dial, and a dark brown leather strap. Watches are the classic product photography torture test. Brushed metal, glass over the dial, stitched leather, tiny textures. If the model survives a watch, it will handle your candle jars and phone cases without complaint.

What is the product shot prompt template?

Google's docs give an explicit recipe for commercial shots, and it's worth quoting because most people never find it: a studio-lit product photograph of the product on a described surface, a named lighting setup, a stated camera angle, and a closing instruction about where the focus should be sharpest. In other words, write the brief you would hand a photographer.

Here's the exact prompt behind the image below: "A high-resolution, studio-lit product photograph of a minimalist wristwatch with a brushed steel case, a forest-green sunburst dial with slim silver baton markers and a dark brown leather strap, standing upright on a low light-gray stone pedestal against a seamless off-white background. The lighting is a three-point softbox setup that creates soft gradients across the brushed steel and a gentle shadow under the strap. Slightly elevated 45-degree camera angle. Ultra-realistic, with sharp focus on the sunburst texture of the dial."

Notice what came back without being asked. The hands sit at roughly ten past ten, the position nearly every real watch ad uses because it frames the logo. The model learned product photography conventions from product photography, so when your prompt sounds like a brief, you get the conventions for free.

The parts that did the heavy lifting were the surface ("low light-gray stone pedestal") and the light ("three-point softbox setup"). Leave those out and you'll still get a watch, just on a random table under random light, different on every run.

Which lighting setups can you name in a prompt?

Lighting is where product shots are won, and Nano Banana Pro responds to the same vocabulary a studio photographer uses. The names that read reliably in my generations: three-point softbox (clean e-commerce look), single hard spotlight (dramatic, deep shadows), soft window light from one side (lifestyle warmth), golden hour backlight (outdoor mood), and overcast daylight (flat and honest, good for color accuracy).

Same watch, opposite brief: "A low-key product photograph... lying on black reflective acrylic with a soft mirror reflection. A single hard spotlight from the top right carves deep shadows, and a faint cool rim light traces the left edge of the case. Dark charcoal background."

Two things in this frame taught me something. The model added faint dust specks on the glass, an imperfection real macro photographers fight with, which reads as realism rather than a flaw. Less welcome: this version grew a fine minute track around the dial edge that the softbox version doesn't have. My product description never mentioned one either way. Between shots, anything you didn't pin down in words is up for renegotiation, which matters the moment you're producing a series instead of a one-off.

How do you shoot your real product, not an imagined one?

An invented watch is fine for a tutorial. Your store sells an actual object, and there are two ways to get it in front of the AI camera.

The first is reference images. According to the Gemini API docs, Nano Banana Pro accepts up to 6 object reference images per request with high fidelity. Photograph your product on your phone from a few angles, decent light, no styling needed, then upload those in the generator and prompt the scene you want around them. The model treats your photos as the ground truth for shape, color, and branding. This is the same mechanism we abuse for keeping faces stable in the character consistency guide; objects are honestly easier than faces.

The second way, the one this article uses, is an exhaustive text description reused word for word across every prompt. It costs nothing and needs no uploads, but it only pins down what you thought to write, as the minute track incident above shows.

One more lesson from the lifestyle shot below, which was prompted as morning light, café table, linen napkin, coffee, "an open paper notebook":

Look at the notebook. The model stamped it with a real stationery brand's wordmark, unprompted, and filled the page with confident handwriting that dissolves into dream-gibberish when you actually read it. For a blog demo that's a fun artifact. In a paid campaign it's a trademark problem plus an uncanny-valley prop, so proofread every object in frame, not just your product, and be ready to crop, regenerate, or ask for "a blank unbranded notebook". Props lie.

Do the shots survive close-ups and marketplace crops?

Detail shots are where I expected Nano Banana Pro to fall apart, because macro is where physical cameras get expensive. A 100mm macro lens, focus stacking, patience. Here's "an extreme macro photograph of the crown and side of the watch, showing the knurled texture of the steel crown", single key light, shallow depth of field:

The knurling on the crown is sharp and evenly machined, the brushing on the case runs in one consistent direction, and the focus falloff behaves like real glass. I'd ship this frame. For detail rows on a product page, this shot type plus the packshot covers most of what a listing needs.

Formats are handled at generation time rather than in cropping. The docs list ten aspect ratios, and the useful mapping for commerce is: 1:1 for marketplace tiles and catalog grids, 4:5 for feed posts (the café shot above is 4:5), 9:16 for stories and vertical ads, 16:9 or 21:9 for site heroes and banners.

Marketplace main images are the one place the template needs adjusting. Amazon's listing rules, summarized well in Jungle Scout's requirements guide, demand a pure white background (RGB 255,255,255, and their scanner checks), the product filling at least 85% of the frame, at least 1000px on the long side, and no text or logos. My stone pedestal on off-white would fail that scan on two counts. For a main image, prompt "on a pure white seamless studio background, product filling most of the frame" and save the pedestal moods for your brand pages. And the obvious-but-mandatory part: a listing image has to show the item you actually ship. Generated glamour that misrepresents the physical product is a returns machine and a policy violation.

What does an AI product shoot cost?

Current BananaBanana prices for Nano Banana Pro: $0.11 per image at 1K or 2K resolution, $0.20 at 4K. The 2K tier is the quiet bargain, same price as 1K and comfortably above marketplace minimums at 2048px. Every image in this article, cover included, cost about $0.80 total, which is less than the parking near any studio I've rented.

The workflow that keeps it that cheap: iterate composition on the budget tiers, then finish on Pro. Draft at $0.03 on Nano Banana 2 Lite or $0.06 on Nano Banana 2 until the framing and light hold, rerun the final prompt once on Pro at 2K. Prompt structure transfers across tiers almost perfectly; the general craft of writing these prompts is in our Nano Banana prompt guide, and the full price grid is on the pricing page.

Where a physical shoot still wins: exact fidelity to your real product without reference images, guaranteed repeatability across a 40-SKU catalog, and materials the model still fumbles under pressure, like engraved text at macro distance or aggressive mirror-polish reflections. My honest read is that AI covers the packshot-lifestyle-detail trio for a small store completely, and turns into a pre-visualization tool rather than a replacement once a brand needs pixel-identical consistency at catalog scale.

FAQ

Can Nano Banana Pro use photos of my real product?

Yes. The Gemini API docs allow up to 6 object reference images per request, kept with high fidelity. Upload a few phone photos of the product from different angles, then describe the scene, surface, and lighting you want built around it.

What lighting should I ask for in an e-commerce packshot?

"Three-point softbox setup" on a plain background is the standard clean look. For dramatic brand imagery, "single hard spotlight with deep shadows"; for lifestyle, "soft window light from one side". Named setups work; vague words like "professional lighting" don't.

Does AI product photography pass Amazon's image requirements?

The rules are about the image, not how it was made: pure white background (RGB 255,255,255), product at least 85% of the frame, 1000px or larger, no added text or logos, and it must accurately show the item sold. Prompt for a pure white background specifically; stylish off-white fails the automated check.

What resolution should I generate for product listings?

2K (2048px) covers marketplace zoom requirements and costs the same $0.11 as 1K on BananaBanana. Reserve 4K at $0.20 for print or large hero crops.

How much does one AI product photo cost?

On BananaBanana: $0.11 with Nano Banana Pro at 1K or 2K, $0.20 at 4K. Drafts on cheaper tiers run $0.03 to $0.06, so a full packshot-lifestyle-macro set with iterations lands around a dollar.