Testing Nano Banana Pro: A Practical Workflow for Interior Designers

AI
 
 

Most interior design tutorials either stay surface-level or jump straight into jargon. The real question for working designers is much simpler: can this tool actually support a real studio workflow?

Nano Banana Pro dropped this week, and instead of focusing on how “cool” the images look, I wanted to know whether it could respect architecture, understand scale, handle lighting, and sit comfortably inside an existing design process.

So I tested it the same way I test any new tool in my studio:

Can it transform an empty shell into a concept direction?

And can it upgrade an unfinished SketchUp render enough that I could show it to a client?

Below is exactly what happened.

Watch it in action

Resources

  • Nano Banana Pro via the Gemini App or Google AI Studio

  • Photoshop 2026 (Gemini 3 Model)

  • Google AI Studio for parameter controls

  • Design Success Circle for full prompt breakdowns and workflows

Workflow 1: The Empty Shell Transformation

The first test was a classic concept task:

turning a raw industrial shell into a “Parisian Loft” direction while keeping the bones of the space intact.

Think peeling paint, concrete floors, harsh midday light — a great test for how well a model respects existing architecture.

1. The Direct Prompt: Style vs. Structure

I began with the simplest approach:

Upload the empty room → describe the look → let the model take a first pass.

What happened?

It leaned too heavily into “Parisian Loft” and reconstructed the room entirely, replacing windows, shifting proportions, and ignoring key architectural cues.

This is a familiar pattern with AI models:

If the style prompt is stronger than the structural context, the architecture disappears.

2. The Descriptive Pivot: Tell the AI What You See

Here’s where the design process gets interesting.

Instead of only describing what I wanted, I first described what already exists:

  • raw steel-framed windows

  • concrete ceiling

  • open rectangular volume

  • distressed wall texture

Only then did I layer in the new elements: wide plank flooring, soft plaster, Parisian wall paneling.

This flipped the results entirely.

The AI preserved the proportions, window placement, and structural character while applying the new finishes convincingly.

Pro tip:

When you want AI to keep the architecture, describe the existing building before you describe your design.

3. Lighting Tests: Daylight Wins, Night Scenes Lag Behind

I tested two lighting conditions:

Golden Hour

The model handled it beautifully — warm reflections, shadow fall-off, believable depth.

This is where Nano Banana Pro feels strongest.

Night Scene

This was tougher.

It often kept the room brighter than it should be, and shadow behaviour wasn’t always consistent.

Honest note:

Nano Banana Pro handles global lighting well, but still struggles with localized, realistic shadowing at night. Expect several iterations.



Workflow 2: Upgrading a Raw SketchUp Model

For the second test, I used a bare SketchUp elevation of a home gym — the sort of thing you might send to V-Ray if you had more time

The goal:

turn a flat, unfinished export into a concept visual suitable for a direction-setting client meeting.

The Literal Problem: Too Much Copying

First attempt:

I uploaded a reference image of a gym with plywood.

The model didn’t just extract the texture — it tried to rebuild the reference gym inside my SketchUp geometry.

This is where many designers get unexpected results.

When you give the AI a fully composed image, it assumes the whole thing is fair game.

Think of this like a floor plan:

When boundaries are unclear, everything becomes movable
The Fix: Separate Your Inputs

I changed the workflow:

  1. I uploaded the reference image into chat, not the generator.

  2. I asked the model to describe it.

  3. I used that description as the style prompt for the SketchUp render.

This forced Nano Banana Pro to apply the material logic without copying the layout.

The result:

The plywood read correctly, the geometry stayed intact, and the image became a usable concept direction.

Refining Visual Details

Once the base look was working, I pushed it further:

Mirrored Walls

It handled reflections convincingly — but only after I specified what the mirror should reflect.

Joinery and Curves

I tested adding a curved cabinet using a cropped reference.

It wasn’t technically perfect, but it was good enough for early-stage client presentations.

Using Google AI Studio for Precision

If you want tighter control, Google AI Studio offers adjustable parameters:

  • Temperature (how literal the model is)

  • Resolution controls

  • Aspect ratios

This is especially useful for designers preparing client-facing images:

Pro tip:

Lock your aspect ratio to 16:9. It aligns with presentation slides, keeps comparisons consistent, and avoids unexpected cropping.

Summary: Is Nano Banana Pro Ready for Designers?

It’s not flawless — but it’s undeniably useful for early design stages.

Where it works well

  • Daylight and global lighting changes

  • Material replacement on existing geometry

  • Reading SketchUp exports accurately

Where it struggles

  • Night scenes and complex shadows

  • Fine joinery or construction-level accuracy

  • Occasionally adding random objects

Overall, it offers a helpful middle ground between a rough sketch and a polished render.

If you want something fast, flexible, and good enough for concept presentations, Nano Banana Pro is worth exploring.


 

Design Success Circle - JOIN NOW

If you enjoyed this

If you found this helpful, join my free community for templates, tools, and behind-the-scenes workflows. You will find all the prompts listed here

https://www.skool.com/designsuccesscircle/about


 
 

Hi 👋🏽 I’m Joanne!

I’m an interior designer, content creator, educator, and business coach. After studying Economics and Education at uni, the design world beckoned, drawing me to Christie’s in London, where I completed post-grad studies in art & design, and then to Hong Kong, where I founded Eclectic Cool, a design firm and design store. Eclectic Cool represented international brands such as Gubi, &tradition, HAY, Armadillo Rugs & Dinosaur Designs to name a few. My work and store have been featured in Monocle, Conde Nast Traveller, Elle Decor, Expat Living, Cathay Pacific inflight magazines, South China Morning Post, and the ABC (Australia) network and more. I live between on the south coast of Australia and Hong Kong with my husband and cavoodle. I’m the mum of three adult children.

 

Don’t forget to grab the free stuff! For Interior Designers!

Want a free investment guide you can use in your interior design business? 

This 24-page investment guide, made in Canva, is yours to customize with your logo and branding.   Fill out the form to download now!

You might like…

Previous
Previous

Why Interior Designers Need an AI Stack, Not One Tool

Next
Next

Why your projects fall apart — and how intake fixes them