*Scroll all the way down for much better images since the last midjourney update.*
Today I’m playing with midjourney for character and book cover design. You’ve probably seen people posting these to Facebook: what you get is extremely high quality concept art – nearly as good as you might expect to pay thousands of dollars for from a digital artist.
But they’re made by robots, generate instantly, and look amazing. There is some controversy (already) over a few book cover artists and designs selling AI art and concepts – the copyright issues are a little unclear.
But let’s talk about all that later. Here’s some of the images I made, in the last few days.
character design: armor and fabric textures
What is MidJourney and why is it great for authors?
- a cheap, easy to use AI app that can make anything you dream up
- a Discord (chatroom channel) where you post text prompts
- you get 4 samples, and can “upscale” a good one
- or “respin” any of the images to get more like that one
(watch the video to see how it works)
- crazy fun, insanely good, any art or photography style
- great for landscapes and background
- can be great for intricate clothing/armor/character design
- struggles to put everything together (multiple objects, arranged)
- you probably need to make dozens of images before you get a good one
- you can’t *own* the copyright (you can use it for covers, but you can’t stop other people from also using it. Read below for all the details).
Not just for fantasy: works great for realism too
A decade ago, if I need a crashed helicopter in the Vietnam jungle, I’d have to find and buy stock images, photo-mash them together and try to create a scene – it would take days. Now it’s instant, and passably good (needs to be good enough to convey mood, genre and setting, but people won’t look too closely at the small details anyway).
It also does *really* good natural landscapes that look like photographs, and *insanely good* 3D mockups of things that look like actual things you can pick up… like fancy swords or whatnot.
And it’s awesome at historical – since pictures of ancient people and places are hard to come by. Or that classic fantasy, illustrated look.
But I mostly love it for quickly generating characters and scenes – an aesthetic – I’ve aways started my novels with great art, and written stories to match. I just met these characters and locations and I’m eager to write them into a book!
And it can even do tattoos, line art, alchemical illustrations, or anything else… below is an image of a “vampire panda riding a vespa.”
Tips and tricks (getting started with midjourney)
- to start, join a “newbie” group in the left panel. Later you can go into any “general” channel.
- you have to type /imagine to get the right prompt box
- if you lose track of your image in the channel (which always happens) you can find all your images here: https://www.midjourney.com/app you’ll need to be signed in to Discord, and also Midjourney; they’re separate.
Once your images finish generating, you can “upgrade” 1~4 for better quality, or “respin” the best image to see what else it comes up with. I didn’t expect it to figure out “plot bunnies” – though #2 is cute. I’d probably hit U2, V2, and then the “respin all” circle arrow.
- you can respin background scenes, but if you try to respin characters with details or faces, it will degrade with each version
- you can change the resolution or ratio
- It does best with (object) + (scene)… ie “a knight with a sword, in a meadow.” Couples, poses, positions, actions, expressions are all too tricky; with multiple objects it tries to merge them together.
For a simple, illustrated look, try “character design concept.”
Otherwise, here are some prompts to play with:
(noun), standing in a (scene), full body portrait, dramatic lighting, volumetric lighting, symmetrical, Vray, beautiful, gorgeous, ultra HD, epic, cinematic, hyperrealistic, hyper realism, hypereal, sharp focus, raytracing, photorealism, ultra detail, unreal engine 5, octane render, photorealistic
- ultra detailed, octane render, symmetric, majestic, 3D, fantasy, intricate
- hdr+cinematic look + dark light settings + 35 mm film
- scifi, fantasy (genre) or nouveau, pencil line drawing, oil painting, manga (art style)
A few things have to go at the end, with a double dash. Use these to change the orientation:
–ar 9:16 (tall)
–ar 16:9 (wide)
you can use “view looking up at the hill” or “view looking down into the valley” …
There are *tons* more, but that’s enough to get started.
How to use midjourney for book cover art
So here’s the thing, you *might* get a perfect image – great – but that’s 1/2 the cover. You still need excellent, professional typography. I have some tools and templates to help with that, lists of best fonts per genre, etc. But if you also want a paperback version, that’s a whole can of worms. It’s usually best to have a designer help.
But also, given the issues with copyright (below), what I’m excited about is taking certain texture details from a gown or armor, adding a background, fixing details (pose, face, weapons, gleams and lighting, etc). I can do all that in photoshop, but most people can’t. And even if I do all that, I still probably wouldn’t sell the designs or be comfortable charging (mostly because it’s still a sensitive topic, not that there’s anything legally wrong with that, as long as authors understand what went into it).
OR just use it to inspire your writing. It’s great for generating aesthetic moodboards and getting the right “vibe” for your scenes.
OR make an amazing, branded blog with custom featured images for each post that have a shared aesthetic.
OR how about fullpage, interior book illustrations in any style?
Try this: “a full page concept design of an slim crooked haunted house with ravens flying, highly detailed, ink on old worn down paper, slight burns on the edges, trash polka tattoo style, 8k, hyper realistic, unreal engine render, –ar 9:16”
But what about the copyright?
While the latest version of MidJourney gives the creator the copyright, laws around intellectual property and creative IP haven’t been revised to fit AI creation tools yet.
SO using art like this on a book cover is potentially a little risky. For my part, if I use my own Midjourney account and edit pieces together into a book cover design, I don’t think I’m in any danger, but I’d be a little cautious with trying to sell them.
So far I’ve seen a knee-jerk reaction from creatives, and they should be scared. If regular people can make gorgeous illustration and get *exactly* what they want in their minds, why would they hire someone and hope they get it right?
But it’s also true, this is a whole new world that laws aren’t yet established to protect or safeguard. This article for example points out the fact that AI tools learned to draw from consuming millions of images. It’s a “black box” because we can’t even understand the process or see where it’s getting its inspiration from. So it’s impossible to be exactly sure that it isn’t too close for comfort.
Basically, if anybody can reliably and recognizably point out specific features that were copied, then it might be a problem. It’s easy to tell if someone has just spammed a bunch of pictures of famous marvel characters or beloved cartoon icons (protected under copyright law) but less easy if something is “inspired by.”
Even if you put in a specific prompt, like War of the Words of Star Wars or Game of Thrones or Lord of the Rings, you’d probably get a thematic style that is recognizable only in its themes, but not in any specific detail – which is a gray area. Even if people can point out that “this is like” or “this reminds me of” such inspiration isn’t protected by copyright laws.
The other issue which is more confusing, is that currently using MidJourney doesn’t give you the creator any control over your own creative IP. Mostly, MidJourney now has a casual statement that you own all the rights to your image. But in the subtext somewhere (this was pointed out by someone else, I think it’s from midjourney:)
“User, by use of DreamStudio Beta and the Stable Diffusion beta Discord service, acknowledges understanding that such waiver also includes waiver of any such user’s expectation and/or claim to any absolute, unconditional right to reproduce, copy, prepare derivate works, distribute, sell, perform, and/or display, as applicable, and further that any such user acknowledges no authority or right to deny permission to others to do the same with respect to the Content.”
This doesn’t mean you can’t sell the artwork you make or use it commercially. It means you can’t attempt to block anyone else from using it. Especially since MidJourney operates as a kind of shared coworking space or public forum, other people could in theory use your prompts or generated art and then create their own. In my opinion, the chances of someone else using your exact final images is miniscule. But if it happened, you’d have no legal right to get them to take their version or copy down for being too similar to yours.
EDIT: if you’re using the free version, you aren’t allowed to use it for commercial purposes; but if you upgrade to a paid version, then you are. And I believe, you can also make a private channel so other people won’t be seeing your creations.
That’s important to understand, but doesn’t outweigh the exciting benefits or opportunities of using an AI art generating tool for authors. Which is why some cover designers are freaking out about it (and yes of course, AI generated art should be fully labelled as such, and probably be less expensive) – but why authors are already enthralled by its magic. Because it really is magical.
For the first time in history, authors can instantly, immediately, get all their scenes and characters literally painted out for them in seconds. It’s great for idea and scene generation, and potentially for advertisements, promo images, marketing bundles or reader bonuses, and branding a unique and brilliant aesthetic for your stories.
And yes, even book covers. Currently, for the most part, I’d need to also do some photoshopping to make the images useable; though in some cases, adding great typography over AI generated art would be enough – and still be much better than what I, after over a decade designing hundreds of covers, could do on my own. This is because the iterations are so fast. Instead of working on one picture, I can *spin* combinations endlessly until I have something amazing.
Typography is important, so you can’t just cram crappy text on top of AI art, but you could use book design or genre-based design templates to help with that (or my new book cover design tool). In *most* cases, I can see this vastly levelling the playing field, so that hundreds of thousands of authors with zero budget can suddenly get beautiful, illustrated covers for much less money.
It’s beyond us to wrestle with the moral implications or bemoan the fate of art in general. This is happening – this has happened. There’s no putting the genie back in the bottle. Authors, like everyone else, will use whatever tools make it easy to give them what they want. And the early adaptors, in most cases, will flourish.
Which is why, I’m sharing this case study and video of some of the experiments I’ve made using midjourney AI, so you can play with the tool yourself and figure out whether it’s worth it to you. If nothing else, it’s insanely fun. It’s given me some enthusiasm for my projects (I actually even started writing again and got a few thousand new words in my WIP).
As a cover designer, this is obviously bad news, but it’ll probably be months or years (if even then) before most authors are even aware that AI art generators are a thing. I actually like the idea of selling AI generated covers; because it’ll take me less time but I’ll be able to give clients better results – remember, authors are competing against other beautiful covers. There has been one style of photoshop, using the same assets, with handdrawn overlays, that have been popular especially in urban fantasy for the past few years.
The other concern, however, is that soon stock photo sites are probably going to get flooded with AI images; which means cover designers or authors could pay for and download stock images assuming they have the rights, when actually they don’t (although, that’s not exactly an issue, because if you downloaded a stock photo for your cover, you’d have to assume somebody else would use the same image or model anyway). A good cover designer would make the piece “transformative” by adding enough details (less than 10% derivative, but really if *anything* is clearly recognizable, it’s not good). But no author using stock photos could sue another for using the same photo.
Is it legally going to be very messy? Yes. Is it an absolute revolution in technologies that will transform how we create? Also yes.
Isn’t AI just “stealing” the art?
I keep seeing this argument, basically “If the AI tool stole images from 100 sources, how the heck is someone going to license it?” Firstly, I’ve found, even if I want something just like a certain style or pose and add a direct image link in there, the results won’t be anywhere close (not even the same idea or theme). But more generally… if it uses a hundred different sources, that’s not illegal even if I was an artist.
There have been cases of cover designers stealing little tiny bits, a foot, a smile, a knife, when they need it to finish their designs, and those bits are still recognizable – technically I don’t even think that’s illegal but the outrage was awful anyway.
In this case, there won’t be anything recognizable. Here’s an article about how it works. Of course, if you say a command prompt like “Tom Hanks” and get a clear portrait of him, that’s on you, not the tool – don’t do that. And it’s true that someone else might make art with a “Tom Hanks” prompt and you wouldn’t really know. So how do you know that the artist didn’t use somebody’s real face and you just don’t recognize it.
That’s a valid point! It could happen. Just like AI art needs to be disclosed, if I was buying cover art, I’d probably want to see the exact prompts/screenshot of the process as well just to make sure. But since these images are (probably) going to flood stock photo sites making great art cheap and accessible, they’re going to be hard to avoid even if authors are trying to be good.
And also yes, you can’t license it (protect other people from copying or stealing it). I understand the concern, but that doesn’t really bother me personally. For one, it’s just too easy for people to make their own. Two, they’d have to find it and *try* to copy it (which they can’t… the tool won’t recreate it. They’d have to find my exact image without text and steal it. It’s a lot of risk and chance and malice that I just don’t think will ever happen, but sure, it could.
The main thing with a branded cover design series is usually the typography or the layout/arrangement in a unique way: but most genres have recognizable fonts and styles anyway so a lot of them look the same). Three, even if they did steal it exactly on another cover in my genre, it’ll be weird and uncomfortable. People tend to trust the book published first or the more successful book – there are millions of books that never see the light of day.
If I made something cool and somebody else used that art to sell millions of books, sure, I’d be upset. That’s an unlikely hypothetical, but it’s there.
So I totally understand the people saying everything is awful and this is a kraken that cannot be contained. Yes, it is a big deal. New technologies always are. And you’re not wrong: AI couldn’t do this by itself. It’s stealing concept art from illustrators, games and movies – people who have photoshopped things together – to create similar stuff. That’s not cool. But it’s also not technically illegal.*
And it’s also a big deal when artists or designers make great art with no effort, if they aren’t clearly stating how they made it, and it’s completely disruptive, like this CNN post:
+ AI won an art contest, and artists are furious.
It’s also being banned from art websites, which is fair: artist communities are built to share work, and they need to keep their core base happy.
But just for fun, especially for scene creation or generating amazing background descriptions or character descriptions; or perhaps for character art or extra bonuses (posters, prints, business cards, stickers, etc) Midjourney is pretty much mind-blowing; a total revolution in art on a level humanity has never ever seen before.
How much should you pay for AI art
There’s a huge discussion to be had about cost and quality… should something cost more because it’s amazing or because it took the artist a long time to make; not to mention expertise and tools. If it didn’t take the artist long and required no special skills, you probably won’t want to pay nearly as much.
But this argument started during impressionism… people actively hated my favorite artist, William-Adolphe Bouguereau, because he represented perfection of skill and mastery, and his paintings took a long time – it made his paintings valuable but he became a symbol of the gatekeeping artworld.
So then the impressionists and all the new artists who could “paint” even quick impressions of landscapes, couldn’t sell their paintings or be shown in galleries, and hacks like Picasso or Van Gogh had to fight for their work to be recognized – which it is today, equal or greater value – even if it didn’t take any time (like Picasso’s quick sketches that sell for millions).
Obviously this is not *just* the same but you could argue art has been going nowhere for decades, just repeating and copying and rehashing old stuff in new ways. What has value? Recently people with money were investing in computer generated monkeys for millions of dollars. Did that art have value? According to whom? It was bought and sold by people with money…
I’ve seen people selling AI generated art for $100, which I think is a fair price. Obviously it should be clear how the art was made and the limitations of it. Getting just ONE great image often takes hours of “work” (typing out prompts, respinning, getting creative with the phrasing).
And while you *could* just make your own, it’s unlikely you’ll get the exact same thing. That said, I wouldn’t be comfortable spending $100 as a buyer. $20 maybe. But if I was making them, it wouldn’t be worth it for me to sell them for $20, I’d rather just dump them on photo sites.
I’m tempted to make a bunch of premade covers, but I probably won’t, because it’s still such a new and raw issue. Typography is a big part of cover design and even a lot of cover designers/illustrators haven’t mastered it, so premade covers with AI art could range in the hundreds.
You *could* make images you like in MidJourney and then hire someone to photoshop everything together and add text; most designers would balk at that request and refuse. Designers have learned to be careful and safe with where they are sourcing material.
Could you make a better book cover with MidJourney? Against the $350 cover range, maybe. Against the higher levels, $500+ probably not – because those designers usually get the text right. But that’s not the real question. Successful authors who are making money will continue to be safe and buy quality work from professionals.
But the masses of newbie authors with crappy, cheap covers – all the garbage on Wattpad that are already stealing art for covers – all that stuff is about to suddenly get a whole lot better… and that’s scary for the authors who have been winning because they can afford expensive covers.
How to make a book cover (again)
It’s not immediately clear yet how to *use* art like this for a book cover, so I’m going to actually use some of the art above – and some new stuff – to create real, usuable covers, so you can see how to put things together and what it can look like (coming soon!)
Photodump – I’m not even doing the tool justice and it’s getting better fast.
Devil’s advocate… (commercial use)
Something else that’s interesting; I used the prompt “caucasian bar owner, dive bar taiwan” and got these images… which look *just* similar enough to be kind of recognizable to a friend of mine, who owns the “Dive Bar” in Tainan, Taiwan.
If midjourney is looking for pictures of white guys who own bars in Taiwan, that’s only a handful of people; and there’s only one called “dive bar.” So the more specific and realistic you get, the greater the chance you might get something that’s a little too close to comfort, though even here, if you showed it to people who knew him, they probably wouldn’t make the connection immediately – all the background stuff is not quite how his bar looks.
But maybe it’s a fluke. I also tried “realistic self portrait, caucasian foreigner named Derek surrealist painter in Taiwan at an exhibition” and couldn’t get anything that looks like me at all, with or without my specific name… even though I’ve been in the news and media here.
I also got no results from “derek murphy, writing a book” or “derek murphy from oregon, living in taiwan.” Maybe it’s learning not to take likenesses, or I’m not famous enough.
What about Dalle2?
I’ve also played with Dalle2 and made some videos but I’ll summarize here: so far, Dalle2 is smarter but doesn’t look as good. It’s still in learning beta and not really open to the public. So all the pictures kind of work but don’t have that jaw-dropping aesthetic.
However it does two things REALLY well:
1. You can erase just part of an image and redraw it from a prompt; which includes extending the edges.
So if you have part of a character and need her full body; or if you want to change her clothes or the background, you can. But you’re limited to square images right now.
2. Dalle2 is more content aware: I can upload a picture of a a grim reaper holding a girl and it will *try* to recreate the scene. The art sucks, but it’s trying. Midjourney can’t really do this yet – it gets the mood but none of the details. IF/when Dalle2 gets better at the details, being able to say I want something like this… that’s pretty amazing.
Another cool comparison was trying to recreate one of my favorite artworks.
Midjourney did a much prettier version of the same *style* but couldn’t grasp any of the details.
- though this is from the last version, it’s probably better now and I’m sure it’ll be amazing in a few months.
BTW – the above images are just tests. Copying a particular artist’s work is dodgy; even if it creates something entirely new from the original, if the robots are good enough, it may look like a new piece in a series that belongs together. *unless* the artist is long dead (70+ years) and no longer protected by copyright, but even then… you can’t copyright a style, or a concept, or a theme.
A lot of the issues I had with MidJourney just got fixed (and it’s only been a few weeks). So I’m uploading some more photos. In the new test beta which I assume will become a regular feature, faces are *much* better, but so are body parts and figures. It couldn’t handle boobs at all, which led to a massacred top half and twisted arms. These photos almost look photorealistic… I can only assume they’ll get uploaded to stock photo sites and most people won’t be able to tell the difference even if they’re trying to stay away from AI art. I’m particularly impressed with the bedroom pics of Tom Holland, in his bedroom taking a hot selfie.
Some of these use celebrity faces which obviously you don’t want to do. I’m also super impressed with art styles though, in specific anime/comic book/illustrator aesthetics. And the full body poses, which sometimes even mostly get everything right – and the profile/side view which was hard to do earlier.
It’ll be flawless in another month; we are still early days. It’s hard to remind myself sometimes that these photos or these people don’t exist.
Character assets and resources
A cover is usually a character and a background. It has always been very difficult to find full body character poses in a set you can use for a book series; for the last few years, people have been generating assets with Daz studio and then usually replacing the faces and fixing the details. Which is great, but these aren’t exclusive, so many premade cover designers are using the same assets. With Midjourney, I’ve been able to make some pretty amazing character art or assets that are more like high-quality illustrated concepts. They’re too rough still to use straight away, but it’s a nice alternative to either renders or stock photos; and in the near future, they’ll be much better than most other resources available.
Best MidJourney Prompts
If you’re having trouble getting results like this, try these midjourney prompts for a more realistic look:
You can try these, either + or , and —testp at the end for the newer beta which is much better at faces and bodies.
Photograph closeup portrait + 150mm + 8k + UHD + photorealistic + HDR + FStop 2.8 + High octane render + Unreal engine 5 + cinematic + highly detailed –ar 2:3 –testp
DMT Reality rule of thirds, cinematic lighting, 4k epic detailed 4k epic detailed photograph shot on kodak detailed bokeh cinematic hbo dark moody photorealistic –ar 2:3 –testp
Diversity in Fantasy Art
I spent the last few days making mermaid art with Midjourney. One of the things I haven’t heard discussed yet about AI art, is that it will allow diversity in fantasy art by lowering the skill entry: most artists are limited by what they can spend time on, and who is paying for that time.
There’s a current failure of imagination regarding diversity in fantasy because so many people have never seen any alternatives; because it costs too much time and effort to produce them. I made these for no other reason than so that when people search for mermaid fantasy art, they won’t see only the “traditional” (white-skinned) mermaids they’re used to.
So I made a bunch of black mermaids. What did it take? Time and attention. Previously, possibly, graphic art as a leisure hobby/career choice might have been more restricted to upper/leisure class people and maybe less black artists had the time and training to produce black mermaid artwork; and less buyers would pay big money for black mermaid art because rich art buyers tend to be… well, you get the point. I’m not making that argument here, but it’s an argument I think could be made. AI art removes structural societal issues of accessibility, which may whole new types of art to emerge.
To get just one high quality “black mermaid” illustration, an artist would have to work for days – they might do a handful, but without compensation probably wouldn’t spend months building up a body of work which means: most mermaids will continue to be white; and big changes like Disney’s little mermaid casting a black actress will feel “wrong”.
But there is more in the world than white and wrong.
When the best artists are for hire and out of reach of common public, the art that gets funded is the only art that gets made; artists who sell can afford to focus on improving their craft. Normally I’d say this is great: artists should take the market seriously and focus on making art that has value. That’s the only logical step for artists. That’s why I might have also told an author, not to waste time with DIY covers because they absolutely need a professional cover and it’s worth spending all your publishing budget on the cover, so you can focus on the writing instead and not play around with something you’ll never be able to produce well. But that’s all changing right now.
You need to pay for usage, so I probably spent 3 days + $150 to make all the mermaid images; so even then it depends on who has time and money to invest in creation.
More Book Cover Design Resources for Authors
- Best Book Cover Software, Designers and Services
- How to Design the Best Book Cover in 8 Simple Steps
- Midjourney AI text to image book cover design art
- Custom book cover design (Creativindie)
- Book cover design templates and 3D mockups
- My favorite book cover designers
- where to find images for your cover design
- Best fonts for cover design
PS. I don’t talk about it much, but I started out as an editor, then a book cover designer. I even have a guide to book cover design you can download below, or some free templates. The first two links go over to Kindlepreneur, because his blog has excellent long-form articles. But you can *understand* why your cover matters without really being able to pull it all together yourself.
Cover Design Secrets: free guide
I’ve helped design over 1000 book covers, including hundreds of bestsellers – download my free book to learn all the insider secrets I use to sell more books. Click here to get it now. I’ll also share some of the advanced book marketing tactics I’ve used to make a full-time income with my writing.
*Does Midjourney steal from other artists?
There’s still a lot of controversy around AI art so I’ve written a longer response. The problem I keep seeing people bring up, is that the algorithm is trained on copyright protected art. People don’t want to use these images if they are “sampling” from real artists.
You can use this site: https://haveibeentrained.com to find out if your artwork has been “stolen” by AI tools, and the answer is probably “yes!” – one cover designer commented, “I don’t care if the AI uses a single pixel of my work to generate a piece of art, it’s theft.”
Other people are saying, basically, anybody who defends AI at this point is a villain… because this is proof AI is stealing from artists.
But that’s not quite right. It’s more like going to visit a gallery, or better yet, seeing a paining through a window of a private house. I didn’t take anything – not one pixel – but that single glance might have inspired me to make something a tiny bit similar. I’m not saying it isn’t wrong somehow, in a moral sense.
Sure it is! This sucks for artists.
But it’s not theft if A) nothing has been taken and B) it’s not illegal.
This is how humans make art. If it is illegal for AI to do this – which it might be, but isn’t right now – imagine how that would affect humans? Imagine having everything you’ve ever seen or experienced, all the art and movies and paintings and books, erased from your mind so you couldn’t accidentally sample from it later? Like entertainment you had to immediately get scrubbed once you left the theatre?
AI is new, but fair use laws have been around for a long time, and I don’t think it’s likely that this will be be proven illegal and that AI tools like this are just going to go away.
Is it immoral: it’s obviously controversial. So you’re going to see companies support one side or the other to placate their base and artists using this issue to drive a ton of traffic and visibility. But the majority of every day consumers will use it anyway, even if people close to the art industry who understand the delicate nature of the conflict stay away from it.
As I saw in a YouTube video, this really feels like finding out you’re Harry Potter and can do magic (the REAL fight over AI art); people need to love to use it less than they feel the need to support artists by avoiding it. You can guilt and shame them into hiding it or feeling guilty about it, but if it makes them happy and improves their life, since it is not illegal, you don’t necessarily have the moral upper hand, at least according to Bentham’s greatest happiness principle.
“But it’s wrong!” you say. I understand and validate your emotions and opinions. You can convince me I’m wrong in the comments, with any *new* materials I’ve overlooked.
I know we’re not going to agree on everything, but let’s at least add some nuanced thought fodder.
- artists pull inspiration from everywhere, and it’s fine, as long as they don’t make a recognizable copy. If I need to draw, illustrate or photoshop something, I’ll look for sources, either in my memory or from stuff I find online. Do I need to purchase usage rights for every image I use for inspiration? I do not. It’s legal, for me, because I can hold the images together in my brain and blend them together in a way where I didn’t copy any image directly, but created something completely new. It’s legal for humans to do this. But not (people claim) AI – mostly because we can’t see or understand how the images are made.
- Some of the produced image do have markers like the shutterstock’s white bar with serial numbers or what looks like actual artist signatures. I believe what’s happening here is, it recognizes that “this kind of art” often has “this kind of feature” and creates something similar. So yes, maybe 20% of the massive depository of photos it’s pulling from, are unlicensed shutterstock pictures. But it isn’t using any of those pictures. If it used the picture – any direct part of it – it would need the copyright. But to be trained… it’s a messy process. It would have been better to download high res, purchases images – tens of millions of them – and stored them in private somewhere for the bot to learn. But I’m not dismayed that this didn’t happen, because I don’t believe the bot is breaking any creative rules (bots shouldn’t be held to a higher standards than humans. Unless we don’t believe that they are doing what humans are doing) – creating actual new work that isn’t just a synthesis of everything else out there.
- Stealing from artists directly – this one gets messier. For a lot of fantasy work, there aren’t pictures of let’s say dragons for example. All images of dragons (or mermaids) are illustrations or photoshop creations. However the concept of dragons goes back a long time; and almost all illustrations of dragons are in some way fan art from big game and movie studios, ancient mythology, or derivatives. Nobody knows what a dragon looks like so we all copy and reproduce. So yes, in this case, the AI is “Learning” from artwork that is probably copyrighted in most cases, without permission. But, here’s the thing, I don’t need permission from the artist to be inspired and make something sort of similar that isn’t at all a recognizable copy. That’s what fair use laws are all about. If it’s out there, visible in public somewhere, I can’t help being affected by it, I can’t deliberately avoid it or strike it from the record. Once I’ve seen it, I may be influenced by it and even unconsciously use it in my own creative productions. I will of course be forgiven as long as it isn’t a direct copy (unless someone can look at my art and the source art and say, this is too similar; this must be a copy). It’s my understanding that this can never happen with AI – unless maybe you are actively using particular artists or celebrity’s names and likeness in the word prompts.
- That’s a can of worms; because people are using those terms in search boxes; and AI art is already being uploaded to stock photo sites, which means art that may be too similar for comfort and might be a close enough copy is out there for purchase and buyers won’t know there are any problems with it. Yes that’s a huge deal, and maybe you should be skeptical of buying anything that looks AI from a stock photo site because of this. 1. here’s a case about a famous artist suing because people are using his name as a popular art prompt; that’s a legitimate case. Should their be a huge class-action case, all artists vs. AI tech? Maybe! 2. certain stockphoto sites take on a lot of the risk, so you’re mostly protected even if somebody did upload AI art.
- Moral stances (bots vs artists). I’m seen a number of authors claim they aren’t interested in AI art at all and would rather support artists, and that’s fine. But – here’s the main thing I’ve been telling authors for years – nobody owes you their support. Authors are artists too. We will buy the best art we can afford from the best artist we can find. I personally have spent $15,000 on premade covers and assets that I’ll probably never use because even though it was cool at the time, six months later the quality has gone up. I very liberally supported artists pretending it was fine even if I just got a bit of writing inspiration, and I don’t regret it. BUT I also don’t feel the need to side with artists over the AI. All artists struggle to find something they can make that someone will buy. Book cover art is a neat space with not a ton of competition. Designers may have to focus on selling skills that AI can’t yet handle; or use this controversy to sell their “anti-AI” artwork, which is fine; though I suspect some artists will embrace AI and do very well, because for most consumers, they just want the best art and the best price.
People are claiming the only reason for authors to use AI art is that they are cheap and lazy; or that the legality is unclear; avoiding the critical question of whether the AI art is producing better results than what’s otherwise available.
As a consumer I am not guilty – and shouldn’t be made to feel guilty – for hiring what’s cheap and easy. In fact my only goal should be to hit profit as fast as possible with my books, so I can write more books. But I will absolutely buy amazing cover art that I can’t get anywhere else; regardless of where or how it was made. And again, I’m not advocating for AI book covers at all. I understand there are risks and authors should be very careful about whether or how they source their book cover art.
This is a deep issue with the writing community: many authors believe their book is objectively better and more valuable because it took them so long or because it was so challenging for them to produce. But that’s never the case, and experienced authors usually write more, and write faster, and are more successful – to claim their books aren’t real or good or valuable because they didn’t suffer for their art…. that’s an archaic romantic ideology that does more harm than good.
Likewise quality designers will be able to put out better quality work, faster, at higher prices – or build a team who can. Book formatting used to be something you had to hire out; but new tools like Atticus or Vellum have made it easy enough for authors to DIY. Book covers are complex and most online tools have always been terrible; for example laying on the text on the back and spine. This isn’t something the robots will be able to replicate soon. But right now, accessibility to art generation from text prompts, does seem to threaten illustrators or designers who use photoshop and stock resources to make book cover designs.
And that’s unfortunate, but here’s what I believe: the obsession with supporting artists continues the mythology that art isn’t real work; that artists are superior and deserve to be supported without producing valuable, tangible assets equal or greater to the cash value; that it’s the “regular people” without the skills who must/need to pay for this “specialized skill” EVEN if there is a cheaper or better alternative, which bypasses all the skills required.
This is, photography for artists. This is the telephone, the radio. Countless jobs have been replaced. This job? Creative art? No, never, because we’ve never been here before, we’ve never been close to here before. This is the edge of a final frontier, perhaps. (Though, this isn’t dissimilar to impressionism, when suddenly everybody was outside painting as a hobby and upsetting those who had spent decades practicing their craft at fancy art schools. First they were snubbed and insulted and ignored. Then they became famous and then popular; suddenly people were buying fast work and hating old classics that took a lot of time).
Cover designers telling authors they can’t publish without professional designers; editors telling authors they can’t publish without editing; I’m over the gatekeeping. Indie publishing has often been about working on a budget. It’s terrifying for the publishing industry right now because previously, low-budget homemade books would be immediately recognizable, and this is going to blur those lines in a big way. Readers may have to develop sharper skills instead of just glancing at a cover – to make sure the writing quality is good. But it’s also possible that, in the meantime, great books that weren’t getting any attention because the authors couldn’t get great covers made, might suddenly find a new breath of life.
I’m a philosophy dropout with a PhD in Literature. I covet a cabin full of cats, where I can write fantasy novels to pay for my cake addiction. Sometimes I live in castles.