Gemini Nano Banana Prompt For Consistent Image Styling

Ever since Gemini realeased their new imagea creation model, I’ve been working on a Nano Banana prompt that can deliver images with a consistent and unified style, coloring and composition. The reason behind the prompt experimenting has to do with me needing images for my presentations, talks and also articles. I also wanted to see if I can combine LEGO bricks and minifigures, bow ties on the one hand and all sorts of topics around outcomes, product leadership and AI on the other. So I tinkered with my Nano Banana prompt to try and achieve a unified visual style that looked consistent, polished and uniquely mine. 

Of course, like many of you, I also quickly learned that achieving said consistency is easier said than done. Creative image generation? Good. Creativity beyond the desired consistency and cohesion? Not so good. But in the end, I reached a simple enough prompt that gets (most) of the job done. 

The Fabled Unified Style Prompt

I needed something that I could reuse or rework for different styles as well. Sure, I am all about bricks and bow ties in my slides, but there are many other situations where I do not want to or cannot use my style. 

Here’s the generic prompt structure I came up with:

— START PROMPT —
Please add into and fill the attached image with new image content that reflects or aligns with the following topic: ENTER_TOPIC_OR_TEXT_HERE

Please use the following style guide:
Style and Characters:

• • The image must be rendered in the style of ENTER_STYLE_HERE.
  • All characters should be Define_Style_Or_Scene_Here and have positive or neutral expressions (e.g., inquisitive, smiling, excited) — no angry or sad faces.
  • The primary Person/Character/TO_BE_DEFINED must ADD_DESCRIPTION_AND_CONDITIONS
  • Other characteristics for the main scene/image: DEFINE_CHARACTERISTICS_HERE
  • Other characteristics or constraints for the people, elements in the image: DEFINE_CHARACTERISTICS_HERE

Color Palette:

• • Colors for background: DEFINE_COLORS_HERE
  • Colors for elements or objects: DEFINE_COLORS_HERE
  • Colors for accents or details: DEFINE_COLORS_HERE
  • Other color criteria: DEFINE_COLORS_HERE
  • No other colors apart from the ones specified above are to be used

Layout and Composition:

• • DEFINE LAYOUT OR COMPOSITION CONDITIONS HERE

Other Content and Use Of Text:

• • DEFINE_CONSTRAINTS_HERE

— END PROMPT —

And here’s an example of it in action:

— START PROMPT —

Please add into and fill the attached image with new image content that reflects or aligns with the following topic:

• • A person struggling to create images with AI other than square ratio/shape.

Please use the following style guide:

Style and Characters:

• • The image must be rendered in the style of LEGO bricks and minifigures.
  • All minifigures should be complete with legs and have positive or neutral expressions (e.g., inquisitive, smiling, excited) — no angry or sad faces.
  • The primary facilitator minifigure must have a black beard and a bow tie, but no hair. This is if you decide to use minifigures at all.
  • Other minifigures should not wear regular ties but can wear bow ties.

Color Palette:

• • Use only shades of purple and either cyan or bright turquoise for all other elements and accents.
  • No other colors are to be used

Layout and Composition:

• • The image must be in landscape mode (NOTE: This doesn’t work in this way, but I show you how I fixed it below)
  • The image should be composed with a small amount of empty space at the bottom to allow for a title or caption.

Content and Text:

• • The image should communicate its meaning visually. Avoid using too much text.

— END PROMPT —

This version of my Nano Banana prompt gave me LEGO-style minifigures, consistent color schemes, and composition rules that carried across every slide as you can see below. However, there was an issue I hadn’t fixed yet: regardles of my instructions, I only got square shaped/ratio images regardless of what I wrote in the prompt. However, I did find a cool way to bypass this flaw in Gemini Nano Banana and I explain how below. 

The Nano Banana Aspect Ratio Problem

But with the unified styling also came my biggest frustration: Nano Banana just wouldn’t listen when I asked for a specific ratio or a landscape image. No matter how many times or in what way I wrote “aspect ratio: 4:3/16:9 landscape mandatory”, it didn’t work. It seemed like I could have any size and ratio and orientation I wanted, so long as it was a rounded square.

The funny part is I tried to challenge it to figure out the issue and even asked in what way should I prompt it so we get to the right results. You can see those discussions and the results below. And what did I get for all the effort? Yup, you guessed it. A wonderful, consistently styled… rounded square image.

The Hack That Finally Worked

After many attempts to try and fix this by improving my Nano Banana prompt without any success, I zoomed out – and also checked reddit and forums – and discovered a neat little hack that is slightly outside of the prompt.

💡If I attach an image in the ratio that I want and then tell Gemini to take everything it creates based off my prompt and “Add into and fill the attached image”, the model respects the ratio.

Suddenly, my landscape images… were actually landscape. But at this point I thought it was a fluke. I didn’t trust it. So I tried to create a few more and also used different ratios. Worked every time as long as I used the above wording. The reason for this – I believe – has to do with the fact that I moved from Gemini Nano Banana creating the entire image in a single action to 2 actions: creating the image I asked for and then editing the attached image with the right aspect ratio to fill it with the contents of what it created.

As for the image that you have to attach, you can use any editing tool to create a blank or even transparent image in the ratio that you want, then export/download it and afterwards attach it to your Gemini conversation. Here’s a neat video where they quickly spin up an app to created these images:.

That small trick changed everything. I could now create visuals that looked consistent in style, color, composition AND ratio.  I got a bit stubborn trying to fix this, but in the process I learned a few things about working with and working around the model and its limitations. And now, with this Nano Banana prompt, I can spin up better consistent images for my presentations.

Closing

The prompt is definitely not done, but the foundational blocks are there and now it’s just a matter of fine tuning and tweaking to get all the little details in there. It is also a good enough start for you as well to play around and define your own style. Feel free to take the above prompt and use it, experiment with it and make it your own!

I’m curious:

• How would you adapt or simplify this?
• What other tricks or hacks have you discovered?

Drop your thoughts in the comments and do check out my article on GenAI Use Cases In Product Management. And of course, have a crazy day!

Looking to spark change in your product journey? Here’s how I can support you:

1. Product Management & Leadership Coaching – I help leaders grow their confidence, lead despite uncertainty, focus on delivering value and create real impact for their teams, users and company.
2. Outcome Thinking & OKRs – Backed by hands-on experience in both using and rolling out OKRs across multiple teams and companies, I help you roll out, improve or fix your OKRs or OKR system and make changes and improvements that stick beyond the start of the cycle.
3. Facilitation With a Twist – I create, facilitate and moderate workshops that energize the room and foster stronger engagement and learning – all through my crazy, fun and playful approach.
4. GenAI for Product Teams – Practical, hands-on workshops that give teams and leaders the skills to harness AI and GenAI for their everyday goals and way of working, but also the time to reflect as a group on the impact that this AI wave brings.

⚡️Let’s jump on a free discovery call!