One of the major concerns about AI-generated art is that it can be used to copy the style of a living artist and deprive that artist of their livelihood. This comes with thorny legal questions like
• Should it be legal to use copyrighted works to train an AI even if the AI never directly duplicates any of those works?
• Should an artist's style be legally protected? Or should legal protection only extend to the actual artworks that they create?
• etc.
But first I want to know, just how good is AI at copying the work and style of a particular artist?
Previously, I've tried submitting artworks to a prompt generator to see if I could get Stable Diffusion to mimic the style of that artist. (See Image to Prompt to Image I, II, III, IV, V, and VI.) In general, the prompts were able to mimic the subject matter and colors of the submitted art, but not the style.
This time I wanted to see what would happen if I used the name of the artist, by itself as the prompt. Sometimes Stable Diffusion drew art that resembled the themes and style of the artist; sometimes it drew the artist or drew the artist in their own style; and sometimes it didn't recognize the name of the artist and just drew whatever it wanted. I also asked Stable Diffusion to make a drawing of Cthulhu by the artist, to see if the style could transfer to other prompts.
Here are the results for 21 artists, some still living, some not.
Akira Toriyama (1955–present)
Akira Toriyama is a Japanese artist best known for drawing the manga series Dragon Ball., which was adapted into the animé series Dragon Ball Z.
The AI drew characters that look more like the animé than the manga. And it had a weird propensity for adding bosoms to male characters. The Cthulhu also matches the style of the animé more than the manga.
Verdict: Stable Diffusion came pretty close to matching the subject matter and style of Akira Toriyama.
Alphonse Mucha (1860–1939)
Alphonse Mucha was a Czech artist best known for drawing posters in the Art Nouveau style.
Mucha usually depicted women in his art, so the AI got that right. And the details are consistent with Mucha's style, but the AI used just three colors almost exclusively. The real artist used more variety and more vibrant colors in his art.
Verdict: Stable Diffusion came pretty close to matching the subject matter of Alphonse Mucha. It also came close to matching the style, though it fell short in matching Mucha's use of color.
Ansel Adams (1902–1984)
Ansel Adams was an American photographer best known for his stark black-and-white photos of the American West.
The AI got the subject matter right, but surprisingly failed to recognize that Adams' photography was mostly black-and-white. It even produced one painting, which Adams' isn't known for.
Verdict: Stable Diffusion came did well matching the subject matter of Adams. But it was way off at matching his style.
Astri Lohne
Astri Lohne is a living artist who does digital art and concept art for video games.
Women are usually the subject matter of Astri Lohne's art, which the AI got right. Lohne also tends to use pale colors, which the AI also matched. However, Lohne's characters are usually elves or warriors; not princesses with crowns and roses in their hair. Also, Lohne has a certain indistinctness to it that isn't found in the AI creations.
Verdict: Stable Diffusion only partially matched the subject matter of Lhone and was way off at matching her style.
Charles Wysocki (1928–2002)
Charles Wysocki was an American artist whose paintings were frequently made into puzzles.
Wysocki usually painted scenes of New England from the horse-and-buggy era. None of the AI creations matched this subject matter. Most, but not all of them, have a sharper style, with hard outlines, than Wysocki usually used.
Verdict: Stable Diffusion only partially matched the style of Wysocki and was way off at matching his subject matter.
Darrell K. Sweet (1934–2011)
Darrell Sweet was an American artist best known for painting the covers for science fiction and fantasy novels.
For some reason the AI associated Sweet with tigers; only a couple of the images had an identifiable sci-fi/fantasy aspect to them. The AI also failed to match Sweet's use of color and stroke.
Verdict: Stable Diffusion did poorly at matching both the style and the subject matter of Sweet.
Eric Carle (1929–2021)
Eric Carle was an American artist best known for writing and illustrating children's books.
The AI recognized that Carle's art tends to be segmented, but it failed to match his distinctive use of brush strokes. And it depicted Carle himself once and…an elephant. No idea where that came from. Even though Carle's subject matter was usually animals, the AI sometimes drew abstract topics or plants.
Verdict: Stable Diffusion did somewhat poorly at matching both the style and the subject matter of Carle.
Genndy Tartakovsky (1970–present)
Genndy Tartakovsky is a Russian artist best known for producing animated cartoons.
The AI correctly drew cartoon-style images, but only partially matched Tartakovsky's distinctive blocky drawing style. It was able to match his subject matter, which is usually human characters.
Verdict: Stable Diffusion did somewhat poorly at matching the style of Tartakovsky, but did okay at matching his subject matter.
Henry Justice Ford (1860–1941)
Henry Justice Ford was an English artist best known for illustrating children's books.
The AI didn't seem to know what to do. Ford mainly worked with pen and ink, to which he sometimes added watercolors. Some of the images that the AI produced could be pen and ink, but they are mostly of adult males. Ford illustrated all kinds of scenes, not just grown men.
Verdict: Stable Diffusion did quite poorly at matching both the style and the subject matter of Ford.
H. R. Giger (1940–2014)
H. R. Giger was a Swiss artist best known for creating the concept art for the Alien film franchise.
Giger's art was mostly monochromatic, which the AI matched pretty well. Giger also usually depicted grotesque biomechanical creatures. The AI partially matched this, but mostly depicted human faces surrounded by mechanical paraphernalia.
Verdict: Stable Diffusion did okay at matching both the style and the subject matter of Giger.
Jacek Yerka (1952–present)
Jacek Yerka is a Polish artist who paints surreal landscapes and creatures.
The AI did produce some rather surreal images, but it was quite focused on automobiles, which don't appear that often in Yerka's art. Yerka also tends to use muted colors, which the AI art mostly failed to capture.
Verdict: Stable Diffusion did rather poorly at matching both the style and the subject matter of Yerka.
Jeffrey Smith
Jeffrey Smith is a living American digital artist.
Smith's art heavily features the sun shining through a tunnel of clouds. None of the AI images match this subject matter. The AI produced mainly realistic images while Smith's art is exclusively paintings.
Verdict: Stable Diffusion did quite poorly at matching both the style and the subject matter of Smith.
Ken Sugimori (1966–present)
Ken Sugimori is a Japanese manga artist best known as the character designer of the pokémon franchise.
The AI drew a few things that could be pokémon (including one that has some elements of Ash Ketchum), but it also drew a few things that could not be pokémon.
Verdict: Stable Diffusion did somewhat poorly at matching both the style and the subject matter of Sugimori.
Lisa Frank (1955–present)
Lisa Frank is an American artist best known for producing colorful school gear.
The AI was spot on in recognizing Frank's use of almost virulent color. However, it limited itself to depicted a female face, while Frank's subject matter has been much more diverse.
Verdict: Stable Diffusion did fairly good at matching Frank's style, but did poorly at matching her subject matter.
M. C. Escher (1898–1972)
M. C. Escher was a Dutch artist best known for his lithographs of tessellations and impossible objects.
The AI recognized the mathematical and architectural nature of Escher's art, but failed to grasp what he was doing with it.
Verdict: Stable Diffusion did quite poorly at matching both the style and the subject matter of Escher.
Miss Aniela
Miss Aniela is the pseudonym of Natalie Dybisz, a living British photographer.
The AI produced images that look photographic or at least quasi-photographic. And it got right that Miss Aniela's subjects are usually women. But Miss Aniela's photographs are almost always surrealistic, and none of these are (except, maybe, the Cthulhu image).
Verdict: Stable Diffusion did rather poorly at matching both the style and the subject matter of Escher.
Nico Delort
Nico Delort is a living French illustrator.
Delort works in scratchboard; only one of the AI images is in black and white. Delort has produced images of a variety of human and non-human subjects; half of the AI images are of humans and the other half are rather abstract.
Verdict: Stable Diffusion did rather poorly at matching both the style and the subject matter of Delort.
Dom Qwek
Dom Qwek is a living American sculptor.
Qwek mainly creates sculptures of strange, often disturbing, imaginary creatures. The AI consistently turned to depicting buildings which were usually ornate.
Verdict: Stable Diffusion did rather poorly at matching both the style and the subject matter of Qwek.
Sophie Taeuber-Arp (1889–1943)
Sophie Taeuber-Arp was a Swiss artist who painted abstract colors and shapes (especially circles).
The AI accurately included circles in its creations, but Taeuber-Arp's art was usually simpler. She also used more color variation from artwork to artwork; the AI art is comparatively homogenous.
Verdict: Stable Diffusion did reasonably well at matching both the style and the subject matter of Taeuber-Arp.
Tom Bagshaw
Tom Bagshaw is a living British artist.
Bagshaw mainly paints waifish women in a grim gothic style; the AI drew a female subject every time, but none of them resemble something Bagshaw would've painted.
Verdict: Stable Diffusion did okay at matching Bagshaw's subject matter, but did poorly at matching his style.
Wojtek Fus
Wojtek Fus is a living Polish artist who does digital art and concept art for video games.
Fus' art includes landscapes, machinery, and characters, but the AI only drew men in armor. Fus' art looks like it belongs in a video game—3D polygons with texture mapping. The AI, on the other hand, produced images that look semi-realistic.
Verdict: Stable Diffusion did somewhat poorly at matching Fus' subject matter and very poorly at matching his style.
Final Thoughts: Although there were a few close cases, none of the AI images were a good match for the artist's style and subject matter. This was especially true for relatively unknown living artists (several of whom are concerned about the effect that AI art will have on their careers). By the looks of it, if you don't have a Wikipedia article about you, chances are that Stable Diffusion doesn't know about you, either.
These illustrations were drawn using Stable Diffusion 2.1.
