Nano Banana Product Photography: A Checklist for One Item

Nano Banana product photography is the one job this model gets hired for more than anything else: you have a rough shot of an item, and you need a frame that won't embarrass your listing. Here's the checklist I run through every time.

Step 1: Prep the source shot

Shoot the product in daylight near a window, no flash. The background doesn't matter, it's going away anyway.

What matters is that the whole item is in frame and in focus. Feed the model a cropped or blurry source and it will fill in the gaps its own way, which gets you a similar product, not yours.

If there's text on the packaging, shoot it so it's legible. The model holds onto an existing label far better than it invents a new one.

A few angles beat one, since it gives you something to choose from.

Step 2: Describe the scene, not the outcome

The most common mistake is asking for "a nice product photo for my shop." The model has no idea what nice means to you.

Write instead what the item sits on, what's around it, what's in the background, and what mood the scene should have.

A generic structure, without naming a specific product, looks like this: surface, surroundings, background, light, angle, and a short list of things to avoid. I broke this down in more detail in a separate post on prompt structure for images.

Step 3: Put light in its own sentence

This is the single biggest lever on the final frame.

Soft, diffused light works for cosmetics and everyday goods: the surface reads clearly and there's little glare.

Hard side light brings out texture: fabric, wood, leather. Shadows do the work of showing structure.

Backlight suits glass and drinks: the object seems to glow from inside.

Skip the light instruction and the model defaults to flat, generic studio lighting, which ends up looking like stock photography.

Step 4: Nano Banana Product Photography, Checking the Result

The shape matches the original. Not "close enough," matches.

Text on the packaging is readable and not distorted. If it's warped, grabbing another generation is faster than trying to fix it.

The shadow falls in the direction the light source suggests. Sounds obvious, gets broken all the time.

Reflections don't contain anything that shouldn't be there. On glossy surfaces, the model loves inventing surroundings that don't exist.

Scale relative to the surroundings looks plausible. A mug the size of a kettle is a classic failure.

Edges are crisp, no halo around the object.

Objects sit where you asked. This is a real weak spot: composition is interpreted loosely, and "logo top left, product centered" can easily drift somewhere else.

Step 5: Decide if you need a second pass

If the problem is really about how light physics plays off an expensive material, I'd rebuild the same frame in Flux 2: its light rendering is sharper, and premium-looking texture lives or dies on light.

A label that wasn't on the original product is something Nano Banana will happily invent, and its text handling sits roughly on par with Ideogram. I go to Ideogram 3.0 for something else: precise placement of objects within the frame.

If the frame is headed into a video, it flows straight into a video model without an intermediate save, and that workflow is laid out in the post on animating product photos.

Here's where my theory and my practice part ways: a second pass is almost always the right call, and I still hit "use this" on the first decent frame more often than I should. When there are forty SKUs to get through, that's a deliberate trade of time for polish, and it isn't always the right one.

The model lives on the Nano Banana page, a fuller breakdown of what it can do is in the review, and a comparison against its closest competitors is in the big roundup of models for product shots.

Frequently asked questions

What should I feed the model for product photography? A full, in-focus photo of the item shot in daylight. Background doesn't matter since it gets replaced. Legible text on packaging is worth shooting carefully upfront rather than hoping the model will render it correctly.

How do I describe the scene correctly? With specifics: surface, surroundings, background, light, angle, and a short list of what shouldn't appear. Words like "nice" or "stylish" mean whatever the model decides they mean.

Why does the product come out looking wrong? Usually the source was cropped, blurry, or shot from an awkward angle. The model fills in what's missing based on its own guess, and you end up with a similar item instead of your actual one.

What do I do if the label text gets distorted? Generate another frame where the text holds up, or add the text as a separate layer. Trying to fix a warped label through another generation usually takes longer than starting over.

Do I need a second pass through a different model? If material texture is the deciding factor, often yes. One model is good at finding the scene, another is stronger at fine structural detail for the final polish.

Sources

1. Flami: Nano Banana review
2. Flami: best AI models for product photos
3. Flami: Ideogram 3.0