“At long last!” self-published authors declare, deviously twirling their evil mustaches. “Thank to the power of artificial intelligence, I no longer have to pay a premium for beautiful cover art!”

Okay, I’m embellishing that. Still, it’s hard to ignore the zeitgeist. Last month, social media was inundated with image after beautiful image. Reddit and Facebook briefly became digital art museums of the information age before mods cracked down on the deluge. The situation was not lost on Ars Technica.

The art flood has slowed or stopped, but uncertainty remains. We finally live in a time where something gorgeous and one-of-a-kind can be mass produced faster than toilet paper during an epidemic. Still, why take my word on the matter when you can try it yourself?

• Go to the Midjourney site.
• Click on “Join the Beta.”
• Accept the Discord invite.
• Go to any of the newbie channels.
• Create your art with /imagine <enter your prompts here>

In about two minutes, the bot will reply with four images. You can upscale or choose a few for further processing.

Needing some artwork for an upcoming Degenesis game session, I gave it a shot myself. In my first attempts, the system really struggled to provide variety for my vision. “Rotten man with yellow eyes walking up dark cement stairs.”

Okay… not great. I wasn’t sure why it jumped to a surreal style. I cut out mention of yellow, but the dreamy qualities remained. Perhaps I should have tried something like, “in the style of…” a given artist, or said “realistic.” Here are the results with the corresponding prompts.

• Rotting man walking up dark cement stairs.
• Rotting man walking up dark cement stairs (take 2).
• Shadowy zombine ascending cement stairs (yes, I misspelled that).
• Black background, rotting man climbing stairs, view from top, shadows, dark.

20 images, all rather monotonous. A few different angles and variations, but nothing that stood out cinematically. Why not a picture of the creature emerging from the stairs? A view from directly above, with the figure arching its back to stare at us? Perhaps crawling or leaning on those railings? Why wasn’t the AI suggesting something realistic or Renaissance era? Watercolors or cubism were the default.

“Why didn’t you give it those prompts?” you may ask. I should have, I admit. Routine blinds us.

Frustrated, I decided to try something else. “Dark, underground, operating room, advanced medical devices, 8k.” This was decent, but spartan. I tried again, swapping out the advanced medical devices for “cyberpunk machines.” At a late-night glance, it seemed nice at first. Then I checked again this morning…

The colors are right. It nails that cybernetic future aesthetic. Still, look closer at the nonsense. What medical purpose does that arm extending from the monitor serve? Why is the tube on the right looping so weirdly? Why are the overhead lights on the top left at strange angles? And the other overhead is… melting? Flopping over? Does it need an enhancement pill?

My efforts weren’t a total wash. After much trial and error, one produced piece was good enough to use. I also observed other Discordians producing wondrous fantasy material as I toiled. Just ask the judges of the Colorado State Fair. If someone used that piece as a book cover back in 2018, I’d have sworn the publisher paid $5,000 for the rights…

“So what’s the problem already?” those self-published authors scream. “Gimme! Gimme!”

Well…

1. The Copyright Question

Who even owns the produced art? I’m not a lawyer, not that they’re sure either. A fair starting point is the blog by Holland & Knight, explaining that AI is a tool of the creator. Makes sense… but who exactly is the creator then? The groups who developed the algorithms and services, or the guy who wrote prompt after prompt, finessing the results? The terms of service probably clarify this discrepancy.

Still, there’s more. The magic ingredients include pieces by human artists. Living creators will probably inquire (legally) about any “contributions on their behalf” to the algorithms. Public domain is usually safe, although a precedence exists to revert creations back to private ownership. Would or could those issues invalidate copyright claims on AI-produced works?

In my amateur opinion, these would have been my starting questions. Instead, things took a turn for the crazy a few months ago…

2. Is It Even Copyrightable?

The usual legal assumptions jumped the rails back in February, when the U.S. Copyright Office claimed that AI artwork “lacks the human authorship necessary to support a copyright claim.”

I’m serious. Here’s the document from the government site itself. Taken at face value, some might assume…

1. No one owns it, so anyone can use it no one can legally stop you.
2. If Amazon or other printers have doubts, they will reject your cover.

Even if the above points are true, it also means that AI generated work cannot be protected. Imagine if you create a fantastic book, and use AI to produce a jacket. Let’s pretend it sells well and earns critical acclaim. Excellent job! Until someone notices your success and realizes there’s no copyright on the art…

At that point, your legal options would be limited. Trademarking might provide some protections, but that’s not going to stop people outside your industry. Your story is safe from plagiarism, yes, but the art isn’t. Copyright is a double-edged sword.

Book covers are small potatoes though. I’ve heard about graphic novelists and comic book authors producing entire pages with AI. Beware my illustrative friends, be-very-ware…

3. It Can Get Messier

And that’s not all, word nerds. The Copyright Office’s rejection was against an AI creator by the name of Dr. Stephen Thaler. Curiously, he attempted to credit the AI itself as the creator, prompting the ruling in the first place. Thaler feels strongly enough to take the matter to court.

If the legal system feels the U.S. Copyright Office was in the wrong, copyright is going to take some strange new definitions in the coming years. For more thoughts, including moral and ethic questions, check out Kotaku’s article on the subject.

Wrapping It Up

If self-published authors want to use AI for book covers… take the above as a warning. What’s okay today might be overruled tomorrow. Don’t be surprised if you find yourself shopping for a different jacket later, all because of new copyright terms or laws.

Still, I feel better after trying a little prompting. News reports have been promoting fear that we’re going to lose our jobs, etc. After seeing what it does, I’m not terribly worried.

Studios bang through a lot of concept art to nail down what they really want. For artists, I think the true value of AI art is that it quickly establishes the “first image” that people envision. The bar is set once AI shows us what everyone else imagines. If algorithm-generated art becomes the standard, standard is likely all it’ll ever be.

Then it’s up to skilled human artists to give us something extraordinary.