Nano Banana Pro is the “ChatGPT Moment” for Visual Communication

Summary: Google's Nano Banana Pro takes AI image generation mainstream, making professional visual communication accessible to everyone. Integrating with Gemini 3 Pro's reasoning engine, it creates professional infographics, comic strips, and visual resumes instantly. While imperfect, with occasional typos, excessive verbosity,  and generic default styling, it outperforms competitors for practical applications. This advancement will transform how we explain complex information and flood digital media platforms with millions of images every day.

I made all the images in this article with Nano Banana Pro.

The launch of Google’s new image model Nano Banana Pro carries AI images into the mainstream. Now, anyone can create a professional-looking infographic about anything or turn dry material into a comic strip.

Nano Banana Pro is the “ChatGPT moment” for AI image-making. In fact, we can extend the analogy to saying that the original Nano Banana (non-Pro version, released August 2025) and the ChatGPT Image-1 native model (released March 2025) were equivalent to the launch of ChatGPT 3.5 in November 2022: it showed the promise of the technology of having a language model provide the underlying intelligence to art-direct the image model to create good-looking images based on any content the user would care to provide. But just as with ChatGPT 3.5, these version-1 image models had too many errors and were too limited for mainstream use, much as they sparked the imagination.

AI became useful with the launch of ChatGPT 4 in March 2023, and the same can now be said of image generation, which draws on a reasoning model for its underlying intelligence. Nano Banana Pro (launched November 2025) is the model that truly makes AI images useful to a broad audience across business and personal use cases.

Out of the box, Nano Banana has awesome text-rendering capabilities, though it still makes a few typos from time to time. The main advantage of the new model is that it integrates with Google’s new Gemini 3 Pro reasoning model, which makes it far more intelligent than regular diffusion-based image models. It has deep world knowledge and can build up a complex understanding of anything by synthesizing information across web searches.

Yes, this means that your digital content now influences AI artwork. This further means that how your brand, products, and even your personal career are depicted in AI images will be the main way many people think about you in the future.

Of course, Nano Banana Pro knows about all visual styles, but unless you prompt it in some detail, if defaults to render images in a rather generic “AI image” style. As shown below, even a single token added to the prompt can result in quite a stylish image.

I asked for a knolling (a deliberate arrangement of objects at precise right angles) representing UX design in the style of an impressionist painting. More interesting than the standard AI style you get by default.

I am sad to report that Nano Banana Pro was the last straw that led me to cancel my Ideogram subscription. It had its day with better text rendering and prompt adherence than other models in 2024, but those days are past. I’m also close to canceling Midjourney: it still draws more stylish images than Nano Banana Pro, but its prompt adherence is so low that the resulting images, while pretty, don’t depict what I want them to show. Seedream 4 is my current go-to model for pretty images, but Nano Banana Pro will serve most of my needs.

Is it perfect? Absolutely not, and I expect future image models, whether from Google or competitors, to blow Nano Banana Pro out of the water. However, Flux is not it. The new Flux 2 model, launched a few days after Nano Banana Pro, was too little, too late. (Flux 2 would have been a great image model if Nano Banana Pro had not existed.)

One thing Nano Banana Pro can learn from Ideogram is a better user interface for iterative exploration of the latent design space. And, as already mentioned, it could get a sense of style from Midjourney. Google should buy these two services, which will probably be cheap now that they are so outclassed in the main use cases.

Chart Junk or Chart Pull?

My friend Edward Tufte will no doubt deplore the flood of “chart junk” Nano Banana Pro is unleashing on the world. Now that everybody can create good-looking infographics in seconds, people will do so. I’m already starting to see banananized infographics all over social media, and corporate presentations and conference lectures are no doubt not far behind. (See the final section for an example of a slide deck made with Nano Banana Pro in NotebookLM.)

When you look at the infographics I made for this article, you will likely conclude that most of the illustrations do not contribute much to understanding the information. Nano Banana Pro is the first AI image model to create accurate business graphics, but none of my case study examples were based on statistics.

Two years from now, I expect that the next major image model will be integrated with the next major reasoning model, with much higher creativity and capacity for conceptual thinking than what we’re getting now. These combined models will probably be able to design infographics with better images that add insight and understanding beyond mere decoration.

Even though Nano Banana Pro’s infographics use decorative rather than conceptual illustrations, the layout of the information and how the model breaks down complex topics does facilitate understanding. Also, we shouldn’t be so fast to dismiss “mere decoration.” By looking attractive and approachable, these infographics will pull in more users and incite them to engage with the information in a way that would not happen with walls of text or stock photography that might be equally visually attractive but carry no meaning.

Just-in-Time Infographics to Explain Anything

Nano Banana Pro’s greatest strength is its ability to rely on its underlying reasoning engine and Google search to retrieve and synthesize information, presenting it in a visual form as an infographic created just for you and your current question. Here are a few examples.

What does this company do? What are its main products, and what do customers say about them? These common questions are uncommonly difficult to answer on traditional websites, but a custom infographic does the trick. Here, I asked for an infographic about the company MeasuringU. (I already know this company, which is my favorite source for information about quantitative UX methods, but for the sake of the argument, I asked Nano Banana Pro to give me a visual overview.)

The head of MeasuringU, Jeff Sauro, is a good friend, so I asked him what he thought of this AI representation of his company. Here’s Sauro’s reply:

“The AI did a surprisingly good job capturing the essence of our brand, particularly our reputation for Quantifying the User Experience. While we often use 'Measuring' to encompass the substantial amount of qualitative work we do (which is a form of measurement!), the AI correctly identified that our identity is strongly rooted in the quantitative.

In terms of accuracy, there are a few nuances the AI missed. It overlooked Measuring University, which is a core product for us regarding education in the UX community. It also listed our platform simply as 'MUiQ'—we would clarify that as the MUiQ UX Research Platform to give context to those who don’t know the name. Finally, while the content is decent, the visual style missed our brand identity: we avoid all-caps and use white backgrounds with illustrations conveying an Ancient Greek theme. Overall, it’s a certainly a fair snapshot from a 3rd party perspective even if it reduced/missed some of the details of our research services.”

My comment on Sauro’s comments: It’s a good caveat to remember that AI-generated infographics almost always will simplify and thus miss some nuances, so it’s strongly recommended to check the original sources before making any important decisions. As for missing the company’s brand style (including its illustration theme), this would be a major concern if a company were to make infographics for its own use. Some of this can be fixed in prompting, which an outsider wouldn’t need to do.

Another example: summarize a product, here ADPList’s new “AI First Designer” course offering.

I recently saw a social media post recommending two classic books by David Pye. Before spending money to buy them (and time to read them), let’s get infographics to summarize each book:

Since I haven’t read these books yet, I can’t tell whether they faithfully represent all of the most important points in the books. But the infographics definitely give you a quick idea of what they are about.

One book I know inside-out is my own Designing Web Usability, so as another test, I asked Nano Banana Pro to make infographics about it. In this case, I can say that the infographics do represent the book well. (But in the first infographic, I am not sure why it’s showing a drawing that resembles me as the illustration for “User Focus.”)

A final example of requesting an infographic in the moment to give a fast overview of something complicated: Andrej Karpathy recently posted an analysis of the difference between two kinds of intelligence: human (or animal/biological) intelligence and machine intelligence. First, a traditional infographic made without prompting for a specific style, and then one where I prompted for an infographic in the style of a pulp science fiction magazine cover.

I like Karpathy’s analysis, partly cause it aligns somewhat with my own position on the difference between human IQ and AI.

As another benefit of custom-made infographics to satisfy your curiosity, I asked for an infographic to summarize how Karpathy’s theory of the two intelligences has been received by the commentariat:

Your Resume Visualized

Who is this Karpathy fellow anyway? He’s one of the most important (and influential) thinkers today on the meaning of AI and the AI revolution. Among many other insights, he identified “vibe coding” and coined that term.

To give you an idea of Karpathy’s background, I asked for an infographic to visualize his resume:

Impressive, right?

You, too can make your resume into an infographic. You might point Nano Banana Pro at your portfolio website, or if you already have a well-maintained LinkedIn profile, you can use its feature to make that webpage into a PDF file and feed this file to the Banana.

It is highly annoying that LinkedIn will not show profile pages to users who are not logged into the service. This will cut LinkedIn off from being represented in AI results and undermine it in the long run. The PDF file generation is a workaround that should not be needed. LinkedIn needs to get with the program and implement GEO (Generative Engine Optimization).

Talking about GEO, the next level of GEO is probably AI-infographic optimization, to modify your online presence to enhance how you are presented in the kinds of infographics I’m showing here. I have no idea how to do this, but surely AIO (to coin yet another annoying acronym) is the next frontier for optimization consultants.

For now, I strongly recommend that you try the exercise of asking Nano Banana Pro to make infographics about your resume and about your company to see what comes out.

One last example of a career infographic: myself! First, a traditional resume infographic based on, and then one where I asked for the visual to be in the style of a video game. (Both based on the article about my career.)

For the next very short time, there will be some alpha had from including a career infographic with any job application. Hiring managers will be sure to look at this graphic before reading through any boring resume. (In the long term, which may only be a month from now, I expect all job applicants to include a career infographic, so this will change from alpha to table stakes. However, especially in the UX field, there will still be some advantage gained from giving the hiring manager a good infographic, so make a bunch and select the best.)

Storytelling Through Comic Strips

If you’re a faithful reader, you will know that I have been experimenting with comic strips as a way to communicate complex information. See, for example, a set I made in March 2024 with Umesh’s Story Illustrator about directions for the UX profession.

The launch of ChatGPT’s native image model (Image-1) made it easier to create comic strips, with decent text rendering in the speech bubbles and good art direction of the details of each panel, as long as the user specified the overall content and style. (This image model supports a very wide range of illustration styles, though it defaults to a rather generic “AI cartooning style” if you don’t tell it otherwise.) For a range of examples, see my newsletter for May 19, 2025. One downside of the current GPT image model is the lack of character consistency across image generations, which means that it supports multi-page comics poorly. I used a workaround to make 6-page mangas about the need for UX professionals to pivot their careers for AI: generate detailed text descriptions of each character and include this with the prompt for every page where that character should be drawn.

Nano Banana Pro supports multi-page comic strips, as long as you generate all the pages within a single chat session. (This is currently almost impossible, due to Google’s TPU meltdown, which causes frequent generation failures.) Here is a 3-page comic strip I made about how Generative UI will increase demand for UX anthropologists for deep field research into customers’ workflows. Too many words, but otherwise a good comic strip.

Frivolous Images

Here are some examples that you may deem frivolous, but they do have serious uses. For example, making a Chiba image of a person could be used for a birthday invitation or a birthday greeting, and making movie posters of team members could be used for team building or for a launch event. For movie posters, I recommend experimenting with different film genres.

“Create a stylized semi-realistic chibi portrait of a person whose face matches 100% with the reference photo, without altering any facial features. Person a large head and small body, with a soft, natural smile — expressive yet realistic. Person wearing puff yellow jacket and short pant white. Add cool accessories: a leather bracelet, necklace, and sunglasses hanging around his neck for a trendy vibe. Pose: sitting casually on a wooden rocking chair, lifting a glass of Pepsi cola, with a dynamic splash effect frozen mid-air, capturing a lively and cinematic moment. Render realistic details in the skin, fabric, and hair textures, while keeping the chibi proportions cute and expressive. Lighting: cinematic yet soft, combining studio-style contrast with warm highlights for a pleasing, lively tone. Background: clean minimalist, ensuring focus stays on the character. Style: cinematic-cute chibi, semi-realistic texture, expressive tone, dynamic composition, balanced lighting.” (Prompt credit: Shreya Yadav)

“Make a 3×3 grid starting with the 1880s to 2040. In each section, I should appear styled according to that decade (clothing, hairstyle, facial hair, accessories). Use colors, background, & film style accordingly.” (Prompt credit: Blaine Brown) I think the photos in the top row are too generic and don't really reflect how I looked back then, but the later ones are more accurate.

Science fiction movie poster.

Film noir movie poster.

Musical movie poster.

Comedy movie poster.

Varying Visual Extent

By default, Nano Banana Pro is currently too verbose. It’s almost as if it wants to flaunt its text-rendering prowess. However, you can add to the prompt a specification such as “1/4 text and 3/4 illustrations” (for an infographic) or “minimal text” (for a comic strip).

As an example, here is a progression of infographics I made for the same article, varying my prompt from requesting a purely textual infographic to a purely illustration-based infographic with zero text. When you view these infographics in the context I’m supplying here, the no-text option is quite nice, but I don’t think these illustrations would make sense if presented in isolation without having seen the other infographics (with text) first.

The infographics are based on a recent article by  Mario García, titled “What the iconic writers of New Journalism can teach us in the AI era.” I strongly recommend that you read the full article, which is very insightful. However, just as in the previous examples, the infographics give you a short overview of what the article is about, helping you decide whether it’s worth your click.

In case you don’t know Mario García, he is one of the world’s leading newspaper designers who has designed many of the world’s major newspapers (both printed versions and websites). He is also one of the few traditional designers who has successfully made the transition to digital storytelling, both in practice (his recent redesign projects) and in theory (several good books and articles.)

Even though I like García’s article, I disagree somewhat with his conclusion to emphasize sensory impressions as uniquely human storytelling. While correct that current AI only has vision, hearing, plus the ingestion of text, the other senses will follow shortly as part of the robot revolution. And in any case, AI can emulate sensory impressions, even if it only knows them from text, particularly if it’s only generating text, images, and music. (Similar to how AI exhibits empathy that’s often rated as deeper than what humans have, even though AI doesn’t have the same meatware feelings as humans do.)

I believe we should combine human strengths and AI strengths to invent a “new-new” journalism and digital storytelling. (Even if some of these human strengths may gradually also become AI capabilities over the coming decade.)

Even though Nano Banana Pro is much better than previous image models at avoiding errors in its images, here it did draw Gay Talese with three hands.

Slide Decks

You can obviously use Nano Banana Pro “raw” to make individual images for any slide presentation: it supports widescreen aspect ratios, even though I used portrait orientation for the images in this article.

You can also use one of Google’s (confusingly many) other services to produce entire slide decks illustrated with Nann Banana Pro: NotebookLM, which creates individual infographics or full slide decks based on uploaded sources.

Here, I uploaded my article on making a career pivot for the AI age, together with my songs on the demise of UX agencies and how to develop agency. It’s a strength of NotebookLM that it can synthesize information across multiple source documents, but in this case I regret including the songs, because NotebookLM ended up including some of the lyrics on the slides, which seems a bit out of place. (The rhymes are nice when sung, but are odd as slide bullets.)

I made both an infographic and a 10-slide deck. (You can’t specify an exact number of slides, only whether you want an in-depth presentation or an overview.) For both, I requested a Wild West theme for the images, and Nano Banana Pro delivered swimmingly on that style.

Infographic:

Slide deck: