Further Adventures in AI Art

 After playing around more with MidJourney, it finally asked me for (a lot of) money, so I've been investigating the other AI artists and am currently subscribed to NightCafe. 

My task is proving a little more difficult than I anticipated. I've given up on the idea of having a romantic couple kissing. For one thing, AIs don't know a lot about kissing, and tend to mash the noses together in uncomfortable looking ways. I tried getting around that with various cheats, such as Picasso.

And then I started thinking about the cover concept. It's not about the romance, it's about Melina and her angst and anxiety, so after a conversation with my Muses of Inspiration (Amy and Jesse), I've decided to only put one character on the cover -- my main character. 

I was explaining to someone the other day that I don't have clear mental images of my characters. Readers are supposed to make that up. I can write "a handsome man walked into the room" and every single reader is going to be visualizing their own version of a handsome man; if I describe him, most readers will start thinking "hmph, he's not that good looking." 

I realized that in order to have both my characters smooching on the cover, I would have to visualize them, and build digital puppets of them for the AI to move around. I tried visualizing the love interest and completely failed, even though he's pretty easy -- shaved head, sneers a lot, homely-handsome. 

I had much better luck with point-of-view protagonist Melina. I describe her as white-Mexican-Greek, with a mother who is an ambitious social climber. She is short, she is more cute than supermodel-ish, she has dark hair and tan skin. I was playing around with generating faces and came up with an anime girl who is the spitting image of her, especially around the eyes. Eyes that are a little sad, traumatized, been-through-hell.

I tried adding a layer of humanity, which softened her eyes up a little, not too bad. She looks a little more dreamy than sad. I'm not sure if she still looks Latinx or Mediterranean, but she looks like a cute 18 year old who is way over her head.

I tried pushing her a little more towards cartoon too, and got something cute, but I don't think I want it on a book cover. 

My next task, now that I've got a concept of Melina is to create a version of her that includes the top of her head, as well as decide on her artistic style. Anime, photorealistic, in the style of some dead artist? I still like the best of my Midjourney explosions and intend to insert it behind Melina's floating head. 

Conclusion: AI art is definitely a lot like drum machines, in that it's super-complicated and takes a lot of time to master, just like the analog form. You need to have a deft touch with prompt writing, which is a brand new skill opening up for people who want to work as AI interpreters, interviewing creators and helping them visualize whatever's on their mind. 

Right now I'm busy trying to make a new "seed image" or basic version of Melina using my new AI subscription. Once she's saved I can tweak her endlessly and even turn her into a cartoon character. Sort of like The Sims on steroids. Right now she looks like this.