What impact do artistic styles have on #aiart? Part 1
As someone who is still on a steep learning curve when it comes to generative art, I was curious how the AI interprets the styles of different artists when applied to the same prompts. I ran three exercises with three different sets of artists. The first set were artists whose styles were not radically different but also somewhat distinct from each other. For the second set, I picked artists from completely different genres. For the third exercise, I used one artists from the first set, one from the second set and two new artists..
For all three exercises, my base prompt was: “old japanese man, orange accent lighting, iridescent chromatic colour-shifting, dynamic lighting, octane render, 8k, hyper detailed, studio lighting, 3d render, hyper realism, photo realistic man.”
The starting point: “Old Japanese man.”
The first set of artists:
Takehiko Inoue, a Japanese manga artist.
Robin Eley, an Australian hyperrealist painter.
Frank Miller, an American comic book artist.
Edward Hopper, a mid 20th century American realist painter.
From my base image, I asked for variations “in the style of” the each of the four artists. I then upscaled one of the options. Here’s the result of the first upscales.
The most obvious difference is the handling of the orange accent lighting. In the Inoue version it’s on the eyes and ears. With Eley, it’s very heavy on the cheeks and eyes. Miller’s has it on the back of the ears, a gradient on the forehead and a slight tint around the eyes. In the Hopper version, the light is most evenly and naturally applied, as if coming from a high key light.
I think it’s interesting that the Hopper lighting is most natural as his use of dramatic lighting was one of his trademarks. To me, the Inoue version stands out because, not only are the eyes highlighted, but they become a focal point, much like they are with Inoue’s manga heroes. His eyes are always strong and such a distinguishing trait of his characters. It’s not surprising that Miller, who was influenced by manga, also has a character with strong eyes but there’s a distinctly different mood. Miller’s eyes are harder, colder. More suspicous? In fact, the eyes seem to be influenced as much by the artist’s actual eyes, as seen in photos of him, as they are from his artwork.
Here’s the versions after a few more rounds of upscaling without adding new modifiers:
After a couple upscales, the Eley version sees the orange accent lighting now spread more evenly across his face. His eyes, however, have kind of disappeared into his face. With the Hopper version, the AI has abandoned the orange rim light but has really refined the light. However, similar to Eley, we are losing detail on the man’s eyes. I probably like the Miller and Inoue final versions the best. It’s interesting that Miller’s man is no longer stern. His eyes are smiling and he seems more friendly. The Ai has also reinterpreted the orange light on his forehead into a hat. Inoue’s is my favorite. The eyes remain a focal point and the lighting is interesting.
So what does this tell us? For static studio headshots where there is no action in the frame, the influence of different artistic styles is subtle. If you were to show an art critic these images and ask them what are the stylistic influences, I doubt there would be enough here to make an intelligent guess.
For purposes of this non-scientific experiment with a very small sample size, it offers little more than anecdotal observations – that adding the style of an artist to the prompts for a portrait might not do much to create a significant difference.
What happens when we pick four artists from completely different artistic genres? We’ll try to answer that in part 2.
You can receive all my content in your inbox by subscribing to my Substack.
