AI Art: An Author’s Perspective

I follow a few artists on YouTube and have watched numerous videos of them playing with a new tool: computer AI which can create original images. Give the AI a text prompt like, “cyberpunk city crowded with people holding umbrellas during a thunderstorm,” and it will produce images to fit the description. All Artists ask the same question: Will this AI Art tool replace human artists?

 

But that’s not my primary question as I’m not an artist. I’m wondering how this tool can help me as an author.

Clearly, books need cover art, so I wanted to jump at that first. The project I’m working on for NaNoWriMo this year is titled Memory Clear. Creating a cover (whether that will be the final cover or not) helps motivate me to keep writing. I made an account with DALL-E 2 and started feeding it prompts. There’s a bit of a learning curve to find what text you need to type to get the images you expect. I usually use stock photography to make my covers, but the ‘photos’ DALL-E rendered didn’t quite cross the uncanny valley. Since I wanted something for the cover for the NaNoWriMo website, I decided to go with a more painted aesthetic. I gave the AI a super vague prompt, and it made a great image that I hadn’t expected.

I’m happy with this cover for my purposes, but I’m pretty sure I’ll hire a graphic designer specializing in photomanipulation book covers when I publish the novel.

Another area of art the authors will commission is character portraits. I’ve never done them before since artists are expensive. But if an AI could make character portraits for free, I’m down for that. Of course, I ran into the uncanny valley problem again. I also found that the tool struggled with ethnicities. It did okay when I asked it to make a portrait of a woman, but weird things happened if I specified ‘white woman’ or ‘black woman.’

I’m pretty happy with the image I got for Bridget using the prompt ‘photo of athletic older woman with buzz cut.’ Just don’t look too closely at her teeth.

 

Social media is the biggest area where I need more images than I can produce. I knew better than to ask DALL-E for photographs, so I decided I wanted a drawing for my previous post. Specifically, I wanted someone writing on a vintage typewriter in a cluttered bookstore. This prompt gave me only images of men, so I specified a woman writing on a vintage typewriter in a cluttered bookstore. For all the illustrations, there was a severe issue of scale; the typewriters were all huge in comparison to the person using them. The typewriter is still a little large in the image I ultimately chose. Also, it’s floating above the desk. The woman’s face looked odd, but DALL-E will let you erase a portion of the image and it will generate something else there. So, I erased her head, and DALL-E redrew her face from a front view. I decided to use that picture despite its flaws because I had been working on finding the right image for so long.

With limited success in all these areas, I was ready to come to the same conclusion the artists I follow came to: AI art won’t ever replace human artists. However, I realized there was an area where I use tons of images I hadn’t explored yet—my puzzle books.

I recently purchased an indie puzzle book titled Udexia (which is fabulous so far, and I would definitely recommend it). The book has gorgeous illustrations, and I would love to have the same caliber of images in my next puzzle book. My next puzzling escapes book takes place in specific locations, and I wasn’t sure if I could adequately describe a place I had never been. However, a computer would likely know exactly what the inside of the Kunsthistorisches Museum would look like. So I asked DALL-E for a black pen drawing of the inside of the Kunsthistorisches Museum, and it gave me this:

DALL-E’s drawing was exactly what I was looking for, so you’ll likely see a variation of this image in my next puzzle book.

I’m excited to keep playing with AI art and see what else it can do.