AI can now generate architectural floor plans! 🤯

Today, I want to talk to you about generating images with AI…
And even generating architectural plans from a simple prompt.

This is something that wasn’t even imaginable a few months ago.

And yet, it’s now becoming possible thanks to Google’s new AI: Nano Banana, integrated into Gemini.

As we reach the end of 2025, the new AI models — especially the latest version of Gemini — are changing everything.

The revolution of multimodal AIs

But what’s really changing with these new AI models like Nano Banana?

The difference is that it’s the same AI that answers you in text, browses the internet, understands your questions… and generates images.
So it actually has a sense of what a building is, what massing means, what a 3D viewpoint is, etc.
It connects everything on its own and gives you coherent images, even if you’re not using the perfect words.

Before, you had to describe your scene down to the millimeter.
Now, you can just talk to the AI, ask for tweaks, refine things…
It remembers the image it generated and the whole conversation.
It’s as easy as rewriting a text with ChatGPT or Gemini.

And that’s what changes everything.

And honestly, Nano Banana goes even further than ChatGPT here.
Its latest version — released barely a week ago — is still blowing people’s minds.
Everyone is testing use cases left and right…
and what you can do with it is seriously impressive.

An AI that creates images and floor plans

This morning, while scrolling through LinkedIn, I came across a post by Ismaël Seleit — an AI creator I know well — and… big shock.

I hadn’t even imagined this use case: you give a very detailed prompt, you describe your building, and the AI generates a floor plan.

Yes, the plans aren’t perfect and you don’t control every line precisely.

But honestly, the result is clean, coherent, and in under five minutes you can produce a hotel floor plan and test tons of variations.

And the craziest part is what comes next:
you can start talking with the AI to modify the plan, then convert everything into 3D.

Since the AI actually understands the building, it can extrapolate the massing properly.

In two minutes, I had a global view of the building…
and even a 3D view of a hotel room to see what the interior atmosphere could look like.

For me, it’s not a rendering tool — we don’t have full control over the project yet.

But it’s an incredible tool for generating ideas, exploring typologies, and testing directions.

You can start with a simple prompt, or even upload one of your own plans and ask it to turn it into 3D… or modify it.

It’s honestly a whole new playground.

Impact on architectural design

Let me reassure you: we’re definitely not at the stage where AI draws precise plans for you.

Right now, it mostly generates images — not the kind of accurate, technical plans you need for a real project.

We’re still in the realm of inspiration, not executable drawings.
But for getting quick material, it’s amazing.

When you start a project without a clear idea, you can already ask ChatGPT or Gemini to analyze your site, pull out the key information, and suggest directions.

That alone saves a huge amount of time.
Then you generate images — not just one, but several — and you sort through them.

You stay the conductor.
You’re the one deciding what works and what doesn’t.

And once you have a concept you like, getting a realistic image — or even a short video — for your client or your team becomes incredibly easy.
The design cycle speeds up massively.

And above all, you can explore directions you would never have had the time to test before.

A smoother communication with your client

But beyond the design aspect, there’s also communication.
As an architect, you master visual tools that your clients, honestly, don’t always understand.

Very few people know how to read a plan, a section, or even a perspective.

And clients often get swayed by overly polished images from developers, while only architects can see the real issues behind them.

Imagine this: you make a rough sketch that isn’t super clear, you don’t feel like spending hours modeling an option…
and your client rejects it simply because they didn’t understand what you meant.

With generative AI, you can produce a quick view to validate the idea — or at least open a real, productive conversation.

Of course, that doesn’t mean your client can change the project endlessly —
and that’s your job to frame properly.
But it creates a much smoother and more effective back-and-forth.

And then there’s the video aspect, which is incredibly powerful:
with almost no effort, you can communicate your project in a far more immersive way.

It’s a real differentiator — whether you’re in a competition phase, early design… or just testing an idea.

Let’s explore the use cases together

I’ve been an AI trainer for two years now, and even I don’t have time to explore everything that comes out. I test a ton of things — lots of tools, lots of use cases… but clearly not all of them.

And I know how much time it takes. You also have projects, deadlines, an agency to run.

You can’t block one day a week to study AI, and that’s totally normal. That’s exactly why I offer these training sessions: to save you time.

The goal is for you to quickly understand what’s possible, have that “aha” moment,
and when you're back at the office, know exactly what to test and how — without starting from scratch.

We’ll also look at how to interact with AI, how to prompt it, how to correct results.

And when something doesn’t turn out the way you want, we’ll see why:
bad reference image, imperfect prompt, wrong model…

We break all of that down together.

AI-powered agencies have an unfair advantage — now you can level the playing field.

I truly believe that the agencies adopting AI early will, little by little, pull ahead of all the others.

Why?
Because they’ll respond to calls for proposals faster, deliver higher-quality projects, offer more to their clients… all without increasing costs.

They’ll also communicate much more easily: send more images, reassure their clients, get clearer feedback… in short, finally reach a project the client actually wants, and not just “the architect’s project,” which is often a source of stress for them.