Most AI product photography tools have the same fundamental problem.
They generate something that looks like your product. Not your actual product. A plausible AI interpretation of what your product might look like based on your text description.
For brand creators this is not good enough. You need images of your actual product - with its real shape, real label, real material finish, and real color - placed into beautiful AI-generated scenes.
You need images of your actual product - with its real shape, real label, real material finish, and real color - placed into beautiful AI-generated scenes.
Nano Banana solves this problem. It is Google's AI image tool built specifically for multi-reference product placement. You give it your real product image, a scene to place it in, and a description of how it should appear. It generates your actual product placed accurately into that scene.
It is also completely free.
This guide covers everything you need to know to use Nano Banana effectively for professional product photography.
What Nano Banana actually is
Nano Banana is powered by Google's Gemini 2.5 Flash image generation model. It was built specifically to handle multi-reference inputs - meaning you can provide multiple reference images simultaneously and the model uses all of them to inform the output.
This multi-reference capability is what makes it uniquely suited for product photography. Most AI image tools accept one reference image at a time. Nano Banana accepts your product image, your brand logo, a style reference, and a scene image all at once. The model synthesizes all of these references into a single coherent output.
The practical result is product photography that contains your real product - not an AI interpretation of it - in an AI-generated scene that matches your brand aesthetic.
How to access Nano Banana
Nano Banana is accessible through two routes.
Route 1 - Through Higgsfield (recommended for most users): Higgsfield's interface includes direct access to Nano Banana as one of its generation options. This is the most user-friendly way to use the tool. Higgsfield's free tier (150 credits per month) includes Nano Banana access. No separate Google account setup required beyond what you already have.
Route 2 - Through Google AI Studio: Google AI Studio (aistudio.google.com) provides direct access to Gemini 2.5 Flash which powers Nano Banana. This route gives you more control over model parameters but requires more technical comfort. Free with a Google account.
For brand creators focused on product photography: start with Higgsfield. The interface is cleaner, the workflow is more visual, and the free tier is generous enough for regular use.
What you need before you start
Product image with clean background: your real product photographed or rendered with the background removed. PNG format with transparent background is ideal. This does not need to be professional photography - a clean phone photo with good lighting and a white background that you remove with remove.bg or Canva works well.
Quality matters here more than anywhere else in the workflow. A sharp, well-lit, accurately colored product image produces significantly better Nano Banana outputs than a dark, blurry, or color-inaccurate one. Invest 10 to 15 minutes getting a clean reference image before starting.
Scene image (for placement workflow): if you are placing your product into an existing scene - either a Midjourney-generated scene or a real photograph - have that scene image ready. Clean, high resolution, with clear space where the product will go.
Brand logo (optional but recommended): if your product has a visible brand logo, label, or text element that needs to be accurate, upload your logo as a separate PNG reference. This gives Nano Banana a clear reference for the text element it needs to render.
The three Nano Banana workflows
There are three distinct ways to use Nano Banana for product photography. Each produces different results and suits different use cases.
Three workflows, same tool - choose based on control versus speed.
Workflow 1 - Scene placement (most used)
This is the primary Nano Banana workflow. You generate a beautiful empty scene in Midjourney, then use Nano Banana to place your real product into it.
1
Generate your empty scene in Midjourney
Describe the scene without your product - the surface, background, lighting, props, and atmosphere. Leave deliberate space where your product will go. Generate 4 to 6 variations and select the scene with the best lighting and composition for your product.
Example empty scene prompt:
'Minimal product photography scene, empty white marble surface with warm gray veining, soft diffused window light from the left, warm golden undertones, generous negative space in center foreground, clean minimal background with subtle gradient --ar 4:5 --v 8.1 --stylize 150'
2
Open Nano Banana in Higgsfield
Upload your Midjourney scene as the base image. Upload your product PNG as the product reference.
3
Write your placement prompt
Be extremely specific about position, orientation, and how the product interacts with the scene.
Example placement prompt:
'Place the amber glass serum bottle centered in the foreground of the marble scene, sitting directly on the marble surface, bottle vertical and upright, slight 15 degree rotation showing the front label, catching the warm window light from the left with a soft highlight along the right edge of the glass, natural shadow falling to the right on the marble surface, label facing toward camera and readable'
4
Generate
Review the 4 outputs for product accuracy, lighting coherence, and shadow quality. Select the best result.
Workflow 2 - Scene generation with product
Instead of using a pre-generated scene you describe the entire scene including the product placement in one Nano Banana prompt. The model generates both the scene and places your product simultaneously.
When to use this: when you want faster results and are less concerned with the exact scene aesthetic. When the Midjourney scene generation step feels like too many steps for your production volume.
Upload: your product PNG as reference.
Prompt structure: describe the full scene including surface, background, lighting, atmosphere, and then describe exactly where and how the product appears within it.
Example:
'Create a minimal luxury product photography scene with a white marble surface and warm gray veining. Soft diffused natural window light from the left with warm golden undertones. Place the amber glass serum bottle from the reference image centered on the marble surface, vertical and upright, with a slight rotation showing the front label. Natural contact shadow to the right. Clean minimal background. Premium skincare campaign aesthetic.'
The result is less controllable than Workflow 1 because you cannot preview and select the scene before adding your product. But it is faster and often produces excellent results.
Workflow 3 - Multi-reference brand placement
The most powerful Nano Banana workflow. You upload multiple references simultaneously - product, logo, style reference, and sometimes a scene. The model synthesizes all of them into a single output.
When to use this: when your product has a specific logo or label that needs to be accurate and visible. When you have a strong style reference image that defines the exact aesthetic you want. When you need the highest possible brand accuracy in a single generation.
Upload order matters: upload your scene or background reference first, then your product PNG, then your logo PNG, then any style reference.
Prompt structure: reference each uploaded image explicitly in your prompt so the model knows how to use each one.
Example:
'Using the marble scene as the background environment, place the skincare bottle from the product reference image centered on the marble surface. Apply the brand logo from the logo reference to the front of the bottle label accurately and readably. Match the overall aesthetic quality and lighting mood of the style reference image. Warm golden window light, minimal luxury feel, premium campaign quality.'
Writing effective placement prompts
The placement prompt is where most users make mistakes. Vague placement instructions produce inconsistent results. Precise instructions produce accurate placements.
Vague placement instructions produce inconsistent results. Precise instructions produce accurate placements.
Always specify these five elements in your placement prompt:
Position in frame: centered, left third, right third, foreground, background. 'Centered in the foreground' is specific. 'In the scene' is not.
Surface contact: is the product sitting on a surface, floating, held, or elevated? 'Sitting directly on the marble surface with natural contact' is specific. 'On the table' is not.
Orientation and rotation: is the product vertical, horizontal, tilted? At what angle? Which face is toward the camera? 'Vertical and upright, rotated 15 degrees clockwise so the front label faces the camera' is specific. 'Standing up' is not.
Light interaction: how does the light from the scene hit the product? Which side catches the highlight? Where does the shadow fall? 'Catching the warm window light from the left with a soft highlight along the right glass edge, shadow falling to the right on the surface' is specific. 'Good lighting' is not.
Scale reference: how large is the product relative to the scene? 'Occupying approximately one third of the frame height, proportional to the marble surface width' is specific. 'Normal size' is not.
All five elements, every time - vague placement prompts are the most common source of bad results.
Common problems and exact fixes
Product floating above the surface: Add to prompt: 'product sitting directly on the surface, natural contact point visible, realistic contact shadow where product meets surface'
Lighting direction wrong on product: Add to prompt: 'light source from upper left, highlight on left side of product, shadow on right side, consistent with scene lighting direction'
Product too large or too small: Add to prompt: 'product scale proportional to the marble surface, occupying [specific percentage] of frame height'
Label blurry or inaccurate: Upload label as a separate PNG reference. Add to prompt: 'label text from the logo reference rendered accurately and readably on the front face of the product, label facing camera'
Product color looks different from reference: Add to prompt: 'product color exactly matching the reference image, [describe specific color - deep amber glass, warm cognac tones, accurate color rendering]'
Shadow looks pasted or unnatural: Add to prompt: 'natural soft shadow, shadow opacity and direction matching scene lighting, shadow slightly diffused at edges, realistic shadow behavior'
Product shape distorted: Add to prompt: 'product shape and proportions exactly matching the reference image, cylindrical form maintained, no distortion of product geometry'
Getting consistent results across multiple images
For a product catalog or ongoing content program you need your product to look consistent across multiple Nano Banana generations. Here is how to achieve it.
Use the same product reference image every time. Do not switch between different photos of the same product. Pick your cleanest most accurate product reference and use it for every generation.
Save your best placement prompts. When a placement prompt produces excellent results document it exactly as written. Use it as your template for that product in that scene type. The same prompt run with different scenes should produce consistent product placement quality.
Build a scene library. Generate 10 to 15 empty scenes in Midjourney across different styles - marble, wood, concrete, outdoor, interior - and save them. These become your reusable scene library. Running the same product through multiple scenes with the same placement prompt produces a visually consistent catalog.
Keep parameters consistent. In Google AI Studio if you are using that route, note your temperature and other settings when you get good results and replicate them.
Nano Banana versus other product placement approaches
How does Nano Banana compare to the alternatives?
Versus Midjourney Omni Reference: Omni Reference maintains the visual identity of a subject across generations but does not use your actual product image - it creates a Midjourney interpretation of it. Label and text accuracy is inconsistent. Nano Banana uses your actual product image and handles labels significantly better. For products with specific branding Nano Banana wins clearly.
Versus Photoshop manual compositing: Photoshop gives you pixel-level control but requires significant skill and time. A skilled Photoshop composite takes 30 to 90 minutes per image. A Nano Banana generation takes 2 to 5 minutes. Quality at the Nano Banana level is not as precise as expert Photoshop work but for most brand content needs it is more than sufficient and dramatically faster.
Versus traditional product photography: a professional product photoshoot costs $500 to $5,000 per session and takes days from scheduling to delivery. Nano Banana produces comparable quality results in minutes for free. For iterative content production where you need many different scene variations Nano Banana has no realistic traditional alternative at anything close to comparable cost.
A professional product photoshoot costs $500 to $5,000 per session and takes days from scheduling to delivery. Nano Banana produces comparable quality results in minutes for free.
Generate your scene prompts automatically
Before using Nano Banana you need strong Midjourney scene prompts to generate your empty scenes. Our free Product Shot Generator builds these automatically.
Enter your product type, material, brand aesthetic, and target platform. The tool generates 5 Midjourney scene prompts - hero, lifestyle, texture, flat lay, and 3D render - each with the correct aspect ratio and parameters. Use these prompts to generate your empty scenes in Midjourney then bring the best scenes into Nano Banana for product placement.
Want to compare Nano Banana against other product photography approaches for your specific situation? The AI Tool Comparison Wizard recommends the right workflow for your product type and budget.
CropUp Ai