Are “A.I. Studies” A Good Form Of Art Practice? (Plus “Human vs A.I.” Comparison)

2024 Artwork AI Studies article title sketch

Well, I was in the mood for writing about A.I. image generators again and I thought that I’d look at whether or not painting or drawing studies of A.I. images is a good type of art practice. From the rhetorical question in the title, you can probably already guess that I don’t think that it’s the best type of art practice, but this article will focus more on why it isn’t.

And, yes, I’d be lying if I said there wasn’t a certain appeal to the idea of “You’ve trained yourself on us, now let’s train ourselves on YOU!“. Not to mention that if you just want to turn your brain off and paint something purely for the sake of personal practice, then it’s probably an easy option. Likewise, there’s a certain fun and satisfaction from turning A.I. “art” into actual human art. And whenever the A.I. belches out something vaguely cool-looking, your “I have to paint that!” instincts might kick in too.

In fact, the latter of these is what inspired this article. Whilst I’ve painted a few studies of A.I. images for comparisons in articles discussing the technology (like this one) and/or for offline “messing around with my other sketchbook” before, this latest one made me think about how useful this actually is as a form of art practice:

AI Image and human-made practice study (9th January 2024)

This is an A.I. image generated via DreamUp on the 9th January 2024 (with labelling added by me) and a semi-digital practice study that I made of it using waterproof ink pens, watercolour pencils, a waterbrush, a white ink gel pen, a scanner and an open-source graphics program called GIMP.

Whilst painting A.I. studies can be a fun exercise, it isn’t really a good form of art practice. About the only good thing that can be said about it is that, because that A.I. doesn’t always understand things like context or visual storytelling, or even artistic intent, making a study of an A.I. image pretty much forces you to use artistic licence. If you want to add “life” to your painting or make it look like a good painting, then you’re probably going to make all sorts of small changes.

For example, in my human-made study, I emphasised the rain a bit more and played around with the lighting/shading in order to get the sort of atmosphere that I originally wanted the A.I. to create. So, as a way of practicing using artistic licence to improve things, then it is… possibly… useful, I guess.

On the other hand, A.I. images aren’t exactly the most varied or dynamic things. From my own experimentation, unless you directly specify a pose, then it’ll usually default to a dull “just standing there” pose, like in the example. Likewise, unless you directly specify emotions or expressions, then the people will usually just have fairly neutral expressions. It’s all very mediocre. In this regard, it isn’t as good of a type of art practice as just pausing a film and painting what is on screen (eg: the middle of a dramatic moment, with expressions, poses, implied visual storytelling etc...).

There’s also the fact that, and I’m sure I’ve mentioned this in a previous article, in order to use A.I. images as a practice tool you already need to have art experience/skills. At the absolute minimum, you need to know how to copy by sight (eg: looking at the 2D outlines of everything in a 3D scene, judging proportions properly etc..). If you want to use the artistic licence I mentioned earlier, you need to know even more skills – everything from colour theory, to lighting/shading, to composition etc… In other words, it isn’t really suitable for beginners. It might help you to slightly hone or practice skills you already have, but it isn’t great for learning.

Also, it isn’t a great type of art practice because it is basically spoon-fed to you. Instead of actively looking for things to practice drawing or painting, the machine just gives you a selection of images and all you have to do is choose one. Actively looking for inspiration is an important skill for any artist. Whilst traditional practice is different from creating original art, having to actively look for things like photographs, interesting things in real life etc… to study gets you used to the idea to taking a more active role in your art. But, if a machine just gives you something to practice with, then you’ll lose out on this.

Another reason why it isn’t great for learning is that A.I. isn’t as reliable as things like photographs, movie stills etc… when it comes to learning artistic rules. Yes, even within the past year, the technology has improved a lot but it can still make mistakes with fundamental things. I got lucky with my practice study, since the shading on the T-shirt looked vaguely realistic and so did the reflections on the ground but I have no way of knowing if that is exactly what it would look like in real life – even if you can get away with “vaguely realistic” with shading.

Even so, part of why learning from looking at real life, or at photographs is so important is because it teaches you a ton of “rules” which can make your art look better. For example, absolute beginner artists will often just paint water blue. But, if you actually look at water in real life or in photographs, then you’ll realise that – depending on context – it’s either the same colour as the container it is in or the same colour as whatever is above and/or next to it. Yes, if you’re painting water in a blue plastic cup or the mirror-like surface of a body of water (with nothing next to it to reflect) beneath a perfect blue sky, then it will be blue. But, most of the time, water isn’t blue.

As “realistic” as A.I. images can look, they can still be unreliable compared to direct observation or real photographs. As such, it’s a bit more difficult to learn artistic “rules” if you just focus on making studies of A.I. images.

So, no, it isn’t the best form of art practice. It can be a fun exercise, and it’s a good way to practice using artistic licence, but not only do you already need art skills in order to do it, it’s a very “passive” way to find source material, the source material can be fairly bland and it also means that it might be more difficult to learn basic artistic “rules” as well. Still, at the very least, it’s better than doing no practice at all.

————-

Anyway, I hope that this was useful 🙂

Leave a comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.