Recently, I asked Codex to draw an apple in more than 100 different visual styles.
It was only after seeing all those apples side by side that styles like Impressionism, Fauvism, ukiyo-e, and Rococo began to mean something concrete to me. Before that, they were little more than names.
If you cannot picture something, you do not know what to expect. And without a clear expectation, it is hard to tell AI exactly what you prefer, so you end up leaving everything to the model. “Draw a cat” and “Draw a cat in the Fauvist style,” for example, can lead to completely different results.
To get AI to create the kind of image you actually have in mind, you need at least a basic visual vocabulary. But the moment you start exploring the field, you run into a huge number of specialized terms: line, shape, color, composition, perspective, light and shadow, texture, medium, art movement, and so on. You could spend ages going down any one of those rabbit holes.
In the age of AI, digging through resources and learning every basic concept from scratch feels like a cumbersome way to enter an unfamiliar field. Surely AI could put together a personalized starter guide instead.
- Style seemed like the most practical place to begin, so I asked Codex to compile a list of widely recognized visual styles with established names.
- Because the defining works of each style often depict different subjects, it can be hard to tell whether you are seeing a difference in style or simply a difference in content. To keep the subject consistent, I chose an apple: simple, familiar, and instantly recognizable.
That is how I ended up with more than 100 versions of the same apple. I also turned them into a website—the Visual Style Atlas—to help me learn the basics.
The atlas started out as a simple browsing tool. Whenever a style caught my attention, I could open it to get a feel for what it looked like, or shuffle through random styles each day and gradually build some visual intuition.
Then I noticed that some styles looked surprisingly similar at first glance and were easy to mix up. So I added a comparison tool that places two styles side by side, making it easier to spot the differences.
Later, I added a “Guess the Style” quiz to see whether I could actually recognize what I had been looking at. Each round gives you 10 random questions to test yourself.
Of course, this is still only a very basic introduction. I still do not really understand line, color, or composition, and looking at 100-plus apples will not suddenly teach me everything about Impressionism or Fauvism. But styles that once felt completely unfamiliar now leave at least a rough visual impression in my mind.
Getting started is not about memorizing definitions. It is about developing an initial sense of what things look like, then digging deeper when a particular style genuinely interests you. First, expose yourself to plenty of examples. Over time, you can learn how to describe and understand what you are seeing.
If you also struggle to express your visual preferences when working with AI and want a place to start, give the Visual Style Atlas a try.