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 22 artists, some still living, some not.
Albert Bierstadt (1830–1902)
Alan Lee was a German artist best known for his landscape paintings of the American West.
The AI drew landscapes that match well with Bierstadt's interests. But Bierstadt used light and darkness in ways that the AI failed to capture. The AI's use of color is also much more homogenous than Bierdstadt's was.
Verdict: Stable Diffusion came pretty close to matching the subject matter, but not the style of Bierstadt.
Andreas Rocha
Andreas Rocha is a living Portuguese artist who does matte paintings and concept art with sci-fi/fantasy settings.
The AI drew landscapes that match okay with Rocha's interests, but the AI landscapes don't reflect the diversity or breadth of Rocha's art. The AI also produced a generic digital art style that doesn't look anything like Rocha's actual paintings.
Verdict: Stable Diffusion did okay at matching Rocha's subject matter, but not very well at matching his style.
Anton Fadeev
Andreas Rocha is a living Russian artist who does concept art for video games.
The AI clearly had no conception of Fadeev's art. The results look like photographs instead of looking like concept art. None of the results use the same color palette as Fadeev. And none of the paintings are of fantasy landscapes.
Verdict: Stable Diffusion utterly failed to reproduce Fadeev's art style or subject matter.
Bill Watterson (1958–present)
Bill Watterson is an American artist who is best known for his long-running Calvin and Hobbes comic strip.
The AI drew cartoons, which is correct. But the characters are often Watterson himself. The eyes and pen strokes don't match the way that Watterson drew. . The AI also produced a generic digital art style that doesn't look anything like Rocha's actual paintings.
Verdict: Stable Diffusion somewhat poorly at matching both Watterson's subject matter and style.
[Note: The first 10 images just use "Bill Watterson" as the prompt. The the remaining images have "[subject] by Bill Watterson" as the prompt.
Chiho Aoshima (1947–present)
Chiho Aoshima is a living Japanese artist who produces paintings that are a dark blend of surrealism and ukiyo-e.
The AI didn't seem to recognize Aoshima, so it just drew oriental women in whatever style it felt like at the moment.
Verdict: Stable Diffusion somewhat terrible at matching both Aoshima's subject matter and style.
Christopher and Matt Cushman
Christopher and Matt Cushman are living artists who are best known for creating intricate cut-away diagrams of Star Trek spacecraft.
The AI clearly had no conception of the Cushman brothers' art. The results look like photographs (usually of two bearded men) instead of looking like technical drawings. None of the results are cutaways of spaceships.
Verdict: Stable Diffusion utterly failed to reproduce Cushmans' art style or subject matter.
Don Maitz (1953–present)
Don Maitz is an American artist who is best known for painting artwork for the covers of sci-fi and fantasy novels.
Most of the AI's results look like they could appear on the cover of a speculative fiction paperback; but none of them look like they could have been painted by Maitz. And Maitz' work conspicuously features characters which are noticeably lacking in most of the AI artwork.
Verdict: Stable Diffusion did very poorly at matching Maitz' art style and rather poorly at matching his subject matter.
Giuseppe Arcimboldo (1526–1593)
Giuseppe Arcimboldo was an Italian artist who is best known for painting portraits of people made of other objects (like fruit or flowers).
The AI drew portraits. And it drew flowers. But the portrait isn't made of flowers. (The Cthulhu image comes closer to the idea.) The painting style is pretty close, but the AI produced images that are sharper than what Arcimboldo produced.
Verdict: Stable Diffusion did okay at matching Arcimboldo's art style and very poorly at matching his subject matter.
Greg (1939–present) and Tim Hildebrandt (1939–2006)
The Hildebrandt brothers are artists who are best known for painting book covers and posters for sci-fi and fantasy works.
The AI clearly had no conception of the Hildebrandt brothers' art. The results are often abstract and none of them look like they belong on the cover of The Lord of the Rings or on a Star Wars poster.
Verdict: Stable Diffusion utterly failed to reproduce the Hildebrants' art style or subject matter.
Igor Morski
Igor Morski is a living Polish artists who combines fantasy and surrealism in his artwork.
The AI produced some surreal stuff, but most of it features the face of an old Slavic man and it doesn't match the variety of what Morski produces. The AI art looks more realistic and less like Morski's paintings.
Verdict: Stable Diffusion somewhat matched Morski's subject matter, but did rather poorly at matching his style.
Jakub Różalski (1981–present)
Jakub Różalski is a Polish artist who combines bleak realism with speculative elements, like werewolves and giant robots.
The AI produced photorealistic images; Różalski's art is painted. The AI matched the bleakness in Różalski's art, but left out the fantastical elements.
Verdict: Stable Diffusion did rather poorly at matching both Różalski's style and subject matter.
John Blanche (1948–present)
John Blanche is a British artist who is best known for producing line-and-wash illustrations for sci-fi and fantasy publications.
This is one of the better matches, this round. The subjects are mainly fantastical in nature (but they all feature the same old man). And most of them look like a line-and-wash illustration.
Verdict: Stable Diffusion did okay at matching both Blanche's style and subject matter.
Kilian Eng (1982–present)
Kilian Eng is a Swedish artist who is best known for producing intricate color sci-fi illustrations.
The AI produced a bunch of character portraits, which is uncommon for Eng. And the art doesn't have the gritty line work or extreme detail that Eng favors.
Verdict: Stable Diffusion did quite poorly at matching both Eng's style and subject matter.
Marc Adamus
Marc Adamus is an American photographer who specializes in long-exposure landscape photography.
The AI produced photorealistic images which often appear to be long exposures. However, Adamus favors long shots of landscapes while the AI mainly produced closeups of non-landscape subjects.
Verdict: Stable Diffusion did quite okay at matching Adamus' style but poorly at matching his subject matter.
Mercer Mayer (1943–present)
Mercer Mayer is an American artist who is best known as the author and illustrator of the Little Critter books for children.
The AI clearly had no conception of Mayer's art. The results mainly feature human characters that are not depicted in a cartoon style. With the exception of Cthulhu, none of them resemble Little Critter, his family, the mouse, the grasshopper, or the spider (which, sadly, usually only has seven legs—when there's an eighth, it must be a prosthetic).
Verdict: Stable Diffusion utterly failed to reproduce Mayer's art style and subject matter.
Peter Mohrbacher (1983–present)
Peter Mohrbacher is an American artist who does fantasy art, including work for the game Magic: The Gathering.
The AI exclusively produced paintings of women wearing pointy metal tiaras. Only a fraction of Mohrbacher's art features women and none of them have pointy metal tiaras. Mohrbacher's art also has a softer look to it with more creative use of stroke than what the AI produced.
Verdict: Stable Diffusion did rather poorly at matching Mohrbacher's art style and subject matter.
Norman Rockwell (1894–1978)
Normal Rockwell was an American artist who is best known for the sentimental paintings of everyday American life that he produced for The Saturday Evening Post.
The AI did astonishingly well at creating paintings that look like they could have been done by Rockwell. It also picked out appropriate subjects, though it committed the fault of putting Rockwell himself into most of the images.
Verdict: Stable Diffusion did rather well at matching Rockwell's art style and subject matter.
Raphael Lacoste
Raphael Lacoste is a living Canadian artist who does concept art for video games, such as Prince of Persian and Assassin's Creed.
Most of Lacoste's art features tiny human characters in a vast fantasy landscape. The AI matched that but didn't depict as many different kinds of landscape has Lacoste has created. The AI paintings also look sharper than the subtler brush strokes that Lacoste uses.
Verdict: Stable Diffusion did okay at matching Lacoste's art style and subject matter.
Sergey Krasovskiy (1975–present)
Sergey Krasovskiy is a Ukranian painter who specializes in painting dinosaurs and other ancient, extinct creatures.
The AI clearly had no conception of Krasovskiy's art because there are no dinosaurs here. And the images look photorealistic, not like the acrylic paintings that Krasovskiy produces.
Verdict: Stable Diffusion utterly failed to reproduce Mayer's art style and subject matter.
Takashi Murakami (1962–present)
Takashi Murakami is a Japanese artist who is best known for his 'superflat' art theory and art aesthetic.
The AI produced images that are mostly consistent with 'superflat'. But they all feature the same subject whereas Murakami's art covers a variety of subjects.
Verdict: Stable Diffusion fairly well at matching Murakami's style but not very well at matching his subject matter.
Tomasz Alen Kopera (1976–present)
Tomasz Alen Kopera is a Polish painter who makes surreal fantasy art.
The AI clearly had no conception of Kopera's art because it spat out a series images that resemble architectural photography.
Verdict: Stable Diffusion utterly failed to reproduce Kopera's art style and subject matter.
Yoko d'Holbachie (1971–present)
Yoko d'Holbachie is a living Japanese artist who specializes in pop surrealism and lowbrow art.
The AI clearly had no conception of d'Holbachie's art. It just produced a series of photorealistic oriental women. d'Holbachie mainly works in acrylics and depicts bizarre and grotesque creatures.
Verdict: Stable Diffusion utterly failed to reproduce d'Holbachie's art style and subject matter.
Final Thoughts: There were a few cases where the AI matched the human artist pretty well. But for the most part the AI didn't do so hot. 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.
