I have a secret architect for my AI images.

Last week, I posted images like this one on GenevIA's Instagram:

Eye with the Eiffel Tower

And my inbox blew up:

"What are you using?!" "I tried the same thing and got a cyclops eye with a plastic Eiffel Tower 😭*"*

So here it is: today I'm giving you the tutorial.

But first, something that might sting a little:

The problem isn't the tool. It's how you use it.

The tool is Nano Banana Pro, Google's image generator, and it's free (for a good number of images). You and I are using the exact same one. And yet your results look like a Photoshop nightmare.

The difference? I never write my prompts myself. Here's how I do it.

Step Into 2026 With The Complete Intranet Buyer’s Guide

If your employees waste time searching for information or juggling disconnected tools, it’s time for a better approach. Haystack’s 2026 Intranet Buyer’s Guide shows you exactly what to look for in a modern platform, so you can streamline comms and empower your teams.

Download The Free Guide Here

The Architect and the Builder

Nano Banana Pro is the best bricklayer in the world. The details of the iris, the tiny blood vessels, the golden light reflecting in the pupil: it can do it all.

But a bricklayer won't build anything good if all you tell them is "make me something pretty". They need an architect's blueprint.

My architect? Claude (Anthropic's AI).

The 3-step method

1: I brief my architect.

Google published an official prompting guide for Nano Banana Pro. I copy-paste it into Claude:

"Here are the official best practices for Nano Banana Pro. You are now an expert in this model."

- Me

2: I describe my vision. In plain words.

No technical prompt. I talk to Claude the way I'd talk to an art director:

"I want a macro close-up of a blue eye, ultra-realistic. In the reflection: the Eiffel Tower at sunset. We need to see the texture of the iris, the blood vessels, the lashes. The reflection has to be believable."

- Me

Claude generates a structured 15-line prompt: depth of field, lighting, texture, photographic style. Details I would never have thought of.

3: I hand the blueprint to the builder.

I copy the prompt. I paste it into Gemini ("Thinking" mode). Enter.

The result:

Why it works

"An eye with Paris in it" → the model improvises. Flop.

"Macro photography, extreme close-up, deep blue iris, visible stromal fibers, Fresnel reflection of the Eiffel Tower at golden hour, f/2.8, cinematic color grading"

→ studio-quality result.

That second prompt? Claude wrote it. In 8 seconds. From my plain, everyday French.

The architect draws. The builder builds. All I bring is the vision.

Your action plan (10 min)

Search Google for "7 tips Nano Banana Pro", the official guide
Copy it into Claude or ChatGPT
Say: "You're a Nano Banana Pro expert. Help me create perfect prompts."
Describe your image as if you were talking to a friend
Paste the prompt into Gemini (Thinking mode)
Collect the "what's your secret?" messages

The real skill of 2026 isn't mastering one tool. It's knowing how to orchestrate them together.

And now you have the formula.

See you next week 🫡

Salim from GenevIA, in Geneva