Rendering a Vision

A Throwback Article

This was originally published at the Mad Genius Club on January 27, 2024. Following on from last week’s republication of my stance on AI art, this may be helpful to those who do want to use it. And, why human artists will always be in demand. Good, cheap or fast: pick any two. You only get two.

I'd started gathering what I needed for this post, and got a little sidetracked earlier this week. That led to me being in front of the camera on my usual Thursday podcast/livestream appearance, instead of inserting asides in the conversation from, well, the side. Specifically, I was being asked to get ranty about AI¹ art. Which I was happy to do. There are persistent myths, which I fear will never entirely subside, surrounding this particular software tool. I have written about them before, in the Mythos of AI, so this post is more about the other thing that crops up regarding AI.

It's not real art, it's too easy... it's cheating... which really goes back to the cultural perception that being an artist (of any flavor) has to be painful and often fueled by mental illness. Leaving that aside, using the software to render what you want isn't exactly easy. It takes time, and it takes skill, and sometimes? It's not possible. Still, it speeds the process of sketching and rendering up enormously, as Alma pointed out last week, that alone makes it a great tool for artists and designers. When I compare it to the hours I'd spend on poring over stock images trying to find elements that could possibly work for the cover I was designing, it's so much less frustrating than that! Which is not the same as saying that it's simple and there is no more frustration.

What you are seeing is the first set of thumbnails from a piece I was working on. Ignore the crustaceans in the lower corner, they are unrelated (but fun!).

One of the main limitations of AI is that it's not a mind reader (thank goodness) so you need to be able to write out what you want in a coherent and logical (because it's a machine, ergo much more logical than a human) fashion to guide the model in the direction you want. An artist can render out what is in their head, precisely. The software cannot. An artist cannot render out what is in their client's head precisely, unless the client can manage to convey exactly what they want... and most of them cannot.

Continuing that same series of images, refining as I go.

Most artists are not also graphic designers, it's two separate skill sets. I am both - but my artistic style isn't suitable for book covers generally. So I put the designer hat on when I'm working on a book cover. I'm not looking for a precise rendition of my inner vision. I'm trying to create art that is visually gripping, will sell the book, will send genre signals, while also standing out from the pack of other covers on the page. It is not an easy task. It is well worth the time and effort put into it, from the research that begins when I start work on a cover, to the final tweaks to lighting that will enable the cover and text to pop on both thumbnails and print copies.

Shortly after this round, I got something the potential client liked, a lot.

So why am I sharing these images with you? Well, that one's easy. I didn't make the sale. Ultimately, the client wanted a very large set of images, very rapidly, and oh, they didn't want to pay a lot for them. I said fine, you know I use AI, yeah? They agreed, I got a vague art brief to work on, and I started out on the work you see above, which was only for the initial image they requested. The final sketches I sent included one they really, really liked. They asked for small personalization touches. I could do that. Then, they wanted me to make major changes to the image, by hand, and show them all of the work done so they could say that the final work hadn't been done by AI. Can I get this done in less than 24 hours, at the same rate you quoted for AI generated art?

I'm not a starving artist. I have a good sense of my boundaries, after some painful experiences in the past. I said no. If they wanted that to be done just so they could say they weren't using AI images, then that should have been part of the initial discussion. If they expected an artist's handwork, then they need to be prepared to pay for the time the artist put into it. When it's rendered with software, that's exponentially faster, and cheaper. It was a mismatch, and they can settle for good, fast, and cheap, or they can find another path forward. Particularly as they wanted the image the AI spit out, which really didn't need much work (I had already done what postwork was necessary), so what was I supposed to do? Trace over layers so that it could be called handwork?

It's not a huge thing. I shrugged and went on with life. I laughed at the attempted concern trolling elsewhere when someone pointed out that the blog illustrations looked like AI - they are, I said, I don't have time to do up 4-6 illustrations by hand for every post I make. Time is precious. Coming close to losing my husband last year has made me acutely aware that we have so little of it to squander. I value my time highly, and I'm going to set my pricing accordingly. I'm also not going to drop everything for a job that I don't have the time for when it threatens to balloon beyond the initially discussed parameters.

I would like any young artists reading this to know that setting boundaries is what will help you grow and succeed. Not pushing yourself into stress and burnout chasing every dollar. If you feel you need to be faster, then find the tools to help you streamline your process, whether that is using AI for sketches, accounting software that will spit out invoices on command, or something like Trello boards for tracking your available time and what time you spent on a given project. Make sure you aren't paying yourself a pittance of an hourly rate, because that will come back to bite you.

And for artists and authors alike... there are times to render the inner vision in beautiful, precise detail. This is why AI will never replace the human artist. It is becoming much more accurate in rendering anatomy, say, although there are many areas in which it still needs help (and this is why I just laugh when people accuse the software of coping art... if so, then you wouldn't see some of the ludicrous renditions of, say, people in a vehicle.) What the AI cannot be is precise. You may, with hours of work and months of training (both you and the software model!) be able to get something both precise and accurate out of it. It's unlikely. The human artist can have a picture in their head and with sufficient training and skill acquisitions, get that out onto the paper, canvas, or screen. Will it be original? Yes. Will it be derivative? Also yes. Ethics boards frown on raising children in utter isolation to try and create an artist who is wholly free of outside influences.

oh, Midj, honey... and yes I give the AI a nickname. I also have a robot vacuum named Natasha, so deal!

There are also times to consider the ability of the software to buy you time (and time is money, in this business) in helping create great design, which is not, most of the time, a precise rendition of the inner vision. Design is, in many ways, about the vision of the end user. Creating a great book cover is all about design. When I designed the covers for the Sumire trilogy, I had something specific in mind. I knew I'd nailed it when I had a reviewer rhapsodize for a paragraph on the covers, how they had taken them back to their youth reading SciFi, and how the stories lived up to the promise of the cover. I'd managed to create goodwill with my vision, and then backed up that already happy potential reader's vision of what they expected to read. That's successful design.

You can buy them all now!

1. AI is a misnomer. It's certainly not intelligent on it's own, and it's neither sentient nor sapient. The tools we call AI are just that, software tools that create using Gaussian models - quite literally they are building up images from noise.