Most people try to generate product photography by describing their product in a prompt and hoping the AI gets it right.
It rarely does.
The product shape is off. The label is unreadable. The color is wrong. The proportions do not match. You iterate 20 times and the best result is still not something you would put on your website.
This is not a prompting failure. It is a workflow failure.
The best AI product photography in 2026 does not come from a single prompt. It comes from a two-step workflow that uses two different tools for what each one does best. Midjourney generates the perfect scene. Nano Banana places your real product into it.
Midjourney generates the perfect scene. Nano Banana places your real product into it.
The results look like they came from a professional studio. The total cost is $30 per month or less. Here is exactly how to do it.
Why single-prompt product photography fails
When you ask any AI image generator to create a product photo from a text description alone, you are asking it to do something it was not designed for.
AI image generators are trained to create plausible images that match text descriptions. They are not trained to accurately reproduce a specific real-world object they have never seen. Your product - with its exact shape, label design, color, and material - does not exist in Midjourney's training data. So Midjourney invents something that sounds like your description. Close but not accurate.
This is why AI product photography prompts so often produce results that look like a generic version of your product rather than your actual product.
AI product photography prompts so often produce results that look like a generic version of your product rather than your actual product.
The two-step method solves this by separating the two jobs. Job one: generate a beautiful scene. Job two: place the real product into it. Different tools for different jobs.
One prompt guesses at your product - the two-step method places the real one.
What you need before you start
Before running a single prompt you need three things.
A high-quality photo of your product on a plain white or transparent background. This is your product reference image. It does not need to be professional photography - a clean phone photo with good light works. What matters is that the background is removed cleanly and the product is clearly visible.
To remove the background: use remove.bg (free for low resolution, paid for high resolution), Photoshop's Remove Background tool, or Canva's background remover. Export as PNG with transparent background.
A Midjourney subscription. Standard plan at $30 per month is recommended for this workflow because you will be iterating on scenes and need generous generation credits.
Access to Higgsfield. The free tier with 150 credits per month is enough to start. Higgsfield gives you access to Nano Banana (Google's product placement tool) through a simple interface.
1
Generate the perfect empty scene in Midjourney
The first step is creating your product scene without the product in it.
This is counterintuitive. Why generate a scene without the product? Because Midjourney's strength is creating beautiful, photographic, on-brand environments. When you remove the product from the brief Midjourney can focus entirely on what it does best - scene quality, lighting, atmosphere, and composition.
What to describe in your scene prompt:
Surface: what will your product sit on? Marble, wood, linen, concrete, sand, glass, velvet, ceramic tile. Describe the material and its color specifically.
Background: what is behind the surface? Seamless gradient, lifestyle environment, blurred interior, plain colored wall, outdoor setting.
Lighting: this is critical. Describe your light source, position, and quality in detail. The lighting you generate in this step will wrap around your product in step 2 and determine whether the final image looks cohesive or pasted.
Composition: how much space for the product? Where should it sit? Rule of thirds, centered, left frame. Leave deliberate empty space where your product will go.
Atmosphere: the overall mood and color palette. This should match your brand aesthetic.
Example scene prompt for a skincare product:
'minimalist product photography scene, empty white marble surface with warm gray veining, soft diffused natural window light from the left, warm golden undertones, gentle shadow to the right, clean minimal background with subtle gradient, generous negative space in center foreground for product placement, shot on 100mm lens, shallow depth of field, luxury skincare campaign aesthetic --ar 4:5 --v 7 --stylize 150'
Run this prompt and select the scene variation with the best lighting and composition for your product. Upscale it to full resolution (U button in Midjourney). Download the full resolution image.
Tips for better scene generation:
Generate 3 to 4 different scene variations before choosing. Look for scenes where the lighting direction is clear and consistent - this makes the product placement look more natural.
Avoid scenes with strong perspective lines pointing to where the product will go. Flat or slightly elevated angles work best for most product categories.
The empty space in the center or foreground should feel intentional not accidental. Good composition in the empty scene translates to good composition in the final image.
Save multiple scene options. You might use the same scenes for multiple product angles.
Not sure which prompts to use for your product? The Product Description to Shot List tool generates all 5 shot types automatically from your product description.
Once you have your perfect scene, open Higgsfield and navigate to the Nano Banana tool (also accessible directly through Google AI Studio).
The Nano Banana interface accepts multiple reference images simultaneously. This is what makes it uniquely suited for product placement.
How to set up your Nano Banana inputs:
Upload 1 - your Midjourney scene image. This is the environment your product will appear in.
Upload 2 - your product PNG with transparent background. This is what Nano Banana will place into the scene.
Upload 3 (optional) - your brand logo as a separate PNG. If your product has a visible label or logo that needs to be accurate, uploading it separately gives Nano Banana a clearer reference.
Three inputs, one accurate output - the more references you give Nano Banana, the more precise the result.
Writing the placement prompt:
Be extremely specific about where and how the product should appear. Nano Banana reads placement instructions precisely so vague descriptions produce inconsistent results.
Include: exact position in the frame, angle and orientation of the product, how it relates to the surface, any specific details about how it should catch the light.
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'
Run the generation. Nano Banana produces 4 variations. Evaluate each for:
Product accuracy (does it look like your actual product?)
Lighting coherence (does the product lighting match the scene lighting?)
Shadow quality (does the shadow look natural and grounded?)
Label accuracy (if your product has text, is it readable and correct?)
If the first batch is not right, iterate on the placement prompt. The most common issues and fixes:
Product floating above surface: add 'product sitting directly on the surface, natural contact shadow' to the prompt.
Lighting direction wrong: specify the exact light direction explicitly - 'light source from the upper left, highlight on the left side of the product, shadow falling to the lower right'.
Product too large or too small: describe the product scale relative to the scene - 'product occupying approximately one third of the frame height, proportional to the marble surface'.
Label inaccurate: upload the label as a separate reference image alongside the product PNG and reference it explicitly in the prompt.
3
Final adjustments
Once you have a Nano Banana output you are happy with, the image is almost ready. Most results need minor finishing touches.
Common adjustments in Photoshop or Canva:
Shadow refinement: if the shadow looks slightly wrong, add a soft drop shadow layer manually and blend it with the generated shadow.
Color grading: apply a subtle color grade to ensure the product tones match the scene tones perfectly. A slight warm or cool adjustment often helps the product feel more integrated.
Sharpening: apply subtle sharpening to the product to match the sharpness of the surrounding scene if needed.
Cropping: final crop for the specific platform aspect ratio if the composition needs slight adjustment.
These adjustments take 5 to 10 minutes and make a significant difference to the final cohesiveness of the image.
These adjustments take 5 to 10 minutes and make a significant difference to the final cohesiveness of the image.
Applying this workflow to all 5 shot types
The two-step method works for every product shot type. Here is how to adapt it for each:
Hero shot: dramatic scene, product centered, strong directional lighting. Use the most premium scene you generate. This is your main image.
Lifestyle shot: contextual environment with props and setting. Generate a lifestyle scene (coffee table, bathroom shelf, kitchen counter) then place the product into it naturally.
Texture shot: generate an extreme close-up of your product's material surface as the scene (no empty space needed), then use Nano Banana to ensure the product label or branding is accurate in the close-up.
Flat lay: generate a top-down styled scene with props and surfaces, then place the product into the center of the composition from above.
3D render variant: generate a clean gradient background with dramatic studio lighting and no surface, then place the product in a floating centered position.
Get your 5 prompts built automatically
Building scene prompts for all 5 shot types manually for every product takes significant time. Our free Product Shot Generator does it automatically.
Enter your product type, material, colors, platform, and brand aesthetic. The tool generates all 5 Midjourney scene prompts - hero, lifestyle, texture, flat lay, and 3D render - each with the correct aspect ratio and parameters for your platform. It also shows you the exact Nano Banana placement instructions for each shot type.
Once you have your prompts, run the scenes in Midjourney then bring the best ones into Higgsfield for product placement. The whole workflow from prompt to finished image takes 20 to 30 minutes once you have done it a few times.
Want to compare this workflow against other product photography approaches? The AI Tool Comparison Wizard recommends the right workflow for your specific product type and budget.
CropUp Ai