If a T-Rex and a Tiger had a Baby: Working with AI Art
So, as I resurrect this series, I have two elephants in the room. AI art, and AI narration. Let's tackle art fist and leave the narration for a future post.
Use of AI to generate artwork is controversial to say the least. There's a big fear in the artist community that work will be taken away from traditional human created art. I think this fear has merit. There are also those who won't support products with AI art, and that's fair. I used to be on the fence about it, but I've come down on the side that AI art is here to stay, and I'm going to take advantage of it.
For me, it boils down to: I'm not a good enough artist to produce a professional looking book cover, and I'm not taking work away from someone else since I was still doing my own book covers. The art that I'm selling is not the cover of my books; I'm selling the story itself. The book cover is packaging designed to reflect a character, feeling, or essence of the book.
Is AI art easy? I think another misconception about it is that you can describe and get anything you want. It turns out, it's not that easy. It's certainly easier than "getting gud" at drawing, but it's not turnkey. Just like taking a photograph is easier than painting a scene, it doesn't mean all photographs have equal artistic merit. Photography is a real art as well. A camera is a tool.
And so, AI is a tool too. I've had to accept that I will never be a "good" artist. I'm ok as an amateur. I enjoy drawing, and I will continue to do my own original art, even for this book series. They may make the blog from time to time. But they're not sufficient to represent my stories to an audience as book covers.
It's derivative! <gasp> It learned!
So what about the issue of how machines learn? They learn by ingesting copious amounts of data, much of which is on the public internet. I understand the copyright concerns, the "I didn't give it permission to download it and use it for that purpose. It's stealing my work so it can learn to do similar (derivative) things." However, I'm not sure I agree with that line of thinking.
First, the learning aspect: I say to that artist: you're guilty of the same, if guilt it can be called. I don't know a single artist who hasn't studied at the feet of the masters. Many of us used comic books as kids to try to learn to draw. If you go to art school, you study the artists of history, sometimes copying their work. This is true in music too. We're influenced by Bach, Mozart… Muddy Waters, Clapton. Even our writing, our fiction stories…. I have 'orcs' in my world. So does World of Warcraft. So does Dungeons and Dragons. Guess what… "orc" is not taken from folk stories. We all got it from Tolkien! And Tolkien recombined elements of mythology too! His work is derivative in that way, even as it is an original blend and reflection of his heart. Almost all art is derivative in some way, because that's how our own brains work too!
In regards to the copyright issue, you posted your art online. AI is not taking your exact image and re-selling it with a different claim of authorship. If you put something online, you can't tell people not to look at it. Once they look at it, they have copies of it in their brains affecting their thoughts and perceptions. Some of them are artists who might print your image and try to copy and learn from it. I'm not sure I see AI as practically that different in this context.
Granted, I'm not a lawyer nor legislator, and these are my opinions. But I think this view is the way society is going.
Using AI to build a Vision
One of the advantages of a traditional human artist is that you can achieve your exact vision. You get to define everything in the scene. (Granted, if you commission an artist, you lose some of that control too… the artist has a vision and they are not just a skilled substitute for your own hands… and that's also why you hire them.) The AI is very much shaped by the body of work that it learned on. There's a reason if I prompt for a green orc, I get something that's a little too World of Warcraft. (For the image generation, I'm using Microsoft CoPilot, with a subscription).
What I've found is that I can rarely get something that matches what I want. However, I do have some artistic skill, and know my way around Clipart Studio pretty well (which is how I've done all my other digital painting, with that and a trusty Wacom tablet). It started that if I could get 90% close, I could close the gap with my own skill in post-painting and "Photoshopping". Sometimes I would take several generated images, take the elements I wanted, and collage them together, with blending and lighting adjustments to try to make them not look like a bunch of stickers stuck on each other (as best I can). Now, I feel if I can get about 80% of the way there I can bridge the gap.
So, in the case of the imagery selected for these books and this website, the art is in the curation, finishing, and composition of the overall body of work. They are guided by a single vision (the world in my books). For some images I'll go through 50 or 60 versions before I get a handful that I download and decide which one is worth the investment to alter. Sometimes I'm lucky and all I need do is adjust eye color and erase those pesky sixth fingers. Sometimes, like for the cover of Through Rose Colored Goggles, I had to assemble elements from 5 different generated images. This included taking a head from a different image of the man and putting it on that body in one image, and removing the gorgon's hair. (For some reason it can't seem to just do "snake for hair" instead of "snakes in the hair").
I'm willing to accept some difference in vision for the images I select for book covers, or the website's image galleries. Just like letting a human artist do their own interpretation can yield surprises, and be valid experiences of the story. But I do try to get the images close enough that they could be represented in the text itself, and are true to the world. (Did I mention it's hard to get a real gorgon with snakes instead of hair?).
But yes. It's orders of magnitudes faster than traditional art to produce images of this kind, and orders of magnitude cheaper than hiring an artist, and orders of magnitude easier than drawing it myself from scratch. But at the end of the day, I'm an independent author trying to share my stories. I'm not trying to be a visual artist.
There will always be a market…
Will AI replace human art? Certainly not. There will always be a market for other people's art. There will always be that vision that feels more original, more inspired, than something a machine generates. The price of "real" art might go up. I think it will certainly become more competitive. I think that the mid-grade level professional artist who's able to make a living or supplemental income now will have a much harder time in the future. People like me who might spring a grand or two for custom art for a book have other alternatives. (Well, not like me; I was just drawing my own for better or worse).
I think that to make it as an artist in the future, artists will have to 'get gud' even more. I think competition will be tighter, meaning it will be even harder to make a living solely as an artist. (And it's already hard!)
I also think for those of us using AI art for whatever purposes (such as book covers for indy writers), there will also be the risk of blending in with the crowd. It might, however, provide a minimum standard that the average reader will expect, which will increase the pressure to use such tools.
The bottom line, as a friend recently told me: "The scribes were upset when the printing press came along." AI is here to stay. It's going to change things. It's a powerful tool, and we have only just begun to grasp its implications. In the meantime, I have a world to tend.
In the future on this blog, I intend to share examples of the journey building the images. And some of the bloopers along the way. I have one scene where I asked for a tiger and a T-rex, and not only did it give me both of those with an absolutely horrid elf, it also rendered a … well let's just say the tiger and t-rex had a baby.
