Table of Contents
The more I experiment with products in different models, the more I realize this: you still need some basic editing chops to make your product look legit.
But there’s one exception.
This workflow using Nano Banana
With enough attempts, you can create damn-near perfect, high-fidelity product shots.
So clean that a client literally asked if we could resize them and use them on their PDP yesterday. I mean shiiiattttt… time to spin up a new wrapper?
Anyway, in my testing, I’ve found this is where Nano Banana shines, and don’t let anyone tell you otherwise.
The wild part? It’s stupidly easy.
Don’t know what a JSON prompt is? Sweet. Screw Jason anyway I was always more of a Freddie fan
So today, the whole newsletter is dedicated to one thing:
The perfect product fidelity with Nano Banana.
Here’s the exact workflow I use:
Step 1 (optional). Create your reference images with Midjourney 7
I’ll usually grab a photo from Google or a stock site, screenshot it, and reverse-engineer the vibe with image prompt. You don’t really need to do this, but to hedge any minuscule chance of any potential lawsuit, I like to err on the side of caution.
I will then take that prompt and put it into Midjourney 7.
That provides me with a useful background reference:
Image from Getty
Output From Midjourney
Prompt generated from imageprompt.com:
Couple sitting by a campfire, Caucasian woman, mid-20s, and Caucasian man, late 20s, wearing plaid shirts, casual jeans, and sneakers. Smiling and looking at each other, holding metal mugs. Warm campfire with dancing flames in the foreground, casting warm light on the couple's faces. Sandy ground, out-of-focus foliage and tent in the background. Evening light, soft and diffused, with warm tones. Relaxed, intimate atmosphere. Natural outdoor setting, possibly a campsite. Mid-shot composition, eye-level perspective. Detailed textures of the clothing, fire, and sand. Realistic portrayal of a cozy camping scene. Photographic realism, slightly soft focus on the background.Step 2. Gather your product + reference images
Example: Say I wanted Delta cans on top of a Yeti cooler, with fall vibes, in front of a campfire.
Here’s the approach:
Gather the product photos: the cans.
Gather the reference photos: a Yeti cooler, the cozy campfire scene from
Step 1
Product Photos:
Product 1
Product 2
Product 3
Reference Photos:
Generated from Midjourney
Just a good ole Yeti cooler
Step 3. Upload everything into GPT 5 → Get your Nano Prompt
This is where GPT does the heavy lifting:
Upload each product image separately.
Upload the background + surface (if applicable).
In your prompt let GPT 5 know which is which, e.g:
Product: Delta Blood Orange Can
Product: Delta Pink Lemonade Can
Product: Delta Blueberry Açaí Can
Background: Couple sitting by a campfire
Surface: Blue YETI Cooler
Paste the below structured prompt, with the above product, background, and the surface where the product will be sitting leading the prompt:
**[YOUR PRODUCTS + BACKGROUND + SURFACE GO HERE]**
Read and transcribe all visible copy, text, and branding from each uploaded product image.
Create a section called “🟣 GPT 5 Action (reading the images)” where you explicitly list: Each product with the exact copy you extracted
Then generate a section called “🟡 Final Nano Banana Prompt” that I can paste directly into Nano Banana.
When writing the Nano Banana prompt:
Place the products front and center, close to the camera, large and sharp, with all copy unobstructed and perfectly legible.
If a surface or object image is uploaded, place the products naturally on it; otherwise, use a simple neutral base that fits the setting.
Use uploaded background images to set the environment, but render them softly blurred with shallow depth of field so the products and their copy are the focal point.
Match lighting to the background reference while keeping the products glossy, sharp, and text accurate.
Output Format:
🟣 GPT 5 Action (reading the images) → List of extracted copy per product image
🟡 Final Nano Banana Prompt → One clean prompt, structured and ready to paste.GPT then spits back:
🟣 GPT 5 Action → a clean transcription of all the product copy.
Note: Cross-reference this! Make sure it is accurate. This is why this step is included in the output!
🟡 Final Nano Banana Prompt → structured and ready to paste.
Step 4. Upload Your Images To Nano, And Watch the Magic Happen
At ~4¢ a run, I always generate 4 versions and pick the strongest. On my first prompt, these were the two outcomes that worked the best:
Not too shabby
Not too shabby part 2
Step 5. Subtle Tweaks
You’ll almost always want to dial things in.
In the example above, the Yeti logo is stealing the spotlight, and the people in the background should really be drinking Delta.
The fix is simple: re-upload the image into Nano Banana and refine it—like Photoshop, but with text prompts—until it looks the way you want. Note, these are intentional em dashes! Don’t ostracize me!
It takes a little patience, but that’s true of every worthwhile creative process.
This leads to 👉
👉This
Prompt: Remove YETI logo + have the men drinking Delta
Note: I reuploaded the product images here
Step 6. Cheers! Call It A Day
Add final copy or overlays in Photoshop/Canva, or if you want motion, drop into Veo 3 (e.g., in this case, let’s add some fire crackling, subtle product movement, etc.).
Note: The below example is a low-quality GIF of the image, so it renders properly in your email! Yours will look more HD than this 😄
Veo 3 Prompt: Animate the fire and have the men cheersing in the background
So here’s the play this week:
Stop burning hours in design tools just to get usable product shots.
Use Nano Banana for high-fidelity assets, fast.
Layer Veo 3 on top for motion.
This is the first model where I’ve thought: shit, yeah, brands could actually build a product library with this.
Go test it, tweak it, and let me know what you make.
Catch you next week, and have a great weekend! 🍻
Will