Art Hero - Where’s my Plastic Controller?

Remember when the Guitar Hero and Rock Band games were popular? I had decided to capture some of the steps taken when assembling and finishing a picture that started with AI generated content, and had made a point to take some of the intermediary screen captures. The audio in my brain was saying, “See, this is not completely devoid of skill…”

And then I thought of Guitar Hero. I’ve seen some impressive player videos posted online, the classic “OMG that’s amazing” being “Through the Fire and Flames”. My entry to the genre was Rock Band, and there are songs I cannot get through on hard, let alone expert, mode. (I’m looking at you, “Green Grass and High Tides Forever”!) Rock Band inspired me to learn real guitar, and it’s another passion that has stuck with me through the years. But as much as you can look at your buddy blaze through and keep an 8x winning streak in a righteous guitar solo, and appreciate his righteously mad skillz … it’s still not really playing the guitar. Not everyone can do what he did, but at the end of the day it’s still a plastic controller.

So maybe that’s what curating and modifying AI generated art is. Although it takes some level of amateur artistic skill and proficiency with digital drawing tools (I use a Wacom Intuos tablet and Clip Art Studio), it’s kinda like playing Art Hero. It’s fun and rewarding, and I get to pretend. But then I’m a fiction writer so maybe there’s some irony to that … writing is kinda like playing pretend and recording it.

But for my purposes, the AI generation as a starting point allows me to create quality imagery for this site and the book covers.

So let’s get to it then. The first step was determining the scene or idea I wanted to show. The Lightfall page had pictures of Aradma and Arda, but none of Anuit. I have other pictures of Anuit, but she’s young in Lightfall so I wanted a younger looking image. And I decided to do the scene where she is awarded a succubus by the Lords of Dis to be her servitor.

The second step was generating the images over the course of a few days. I use Microsoft CoPilot, and have an active subscription to it since I generate a lot of content. Part of the journey is figuring out how to prompt it. And sometimes you run into content filters… the word ‘succubus’ simply got blocked, so I had to describe Bryona as a proper lady, describe her clothing, hair color and style, and then say she had two tiny devil horns and demon wings. You give it a prompt, and the tool comes back with up to four images. They are often wrong, or it reverses some traits (e.g., puts swaps hairstyles, features, or clothing between the two characters sometimes). But if there’s an image that is close or I can work with, I download it. I tried some images with both Anuit the sorceress and her together, but nothing quite worked, and ended up rendering a bunch of pictures of them individually. And getting characters to face each other, or look away from the ‘camera’ seems really hard.

Like with the above image. There’s nothing wrong with this picture. It’s fine enough.. the changes would be minor (color change on the horns from red to black, and Bryona’s wings are too small, for example). And Anuit’s collar necklace looks weird… I’d probably use the oil brush tool to paint over that to remove it. But other renderings of Bryona had the lace, so is closer to the book descriptions. And there’s not much going on here.

Another example:

Again, nothing wrong with this picture per se, other than Bryona’s eyes need smoothing over… even the hands look ok. In the book scene Anuit is somewhat surprised the Lord of Dis give her a succubus as her tempter servitor rather than in incubus, so I asked the AI tool to make the sorceress look surprised. However, it has a difficult time not applying the same emotion to every subject in the scene, and the succubus Bryona certainly wouldn’t be surprised. When this came up, I laughed. Plus her wings are small (I could paint them larger), and her dress is farther away than what I asked for, compared to some of the others rendered.

I started focusing on rendering the individual characters separately, with a plan to assemble the final scene.

One of the Anuit downloads:

While this image is stunning, and she has the correct number of fingers, Anuit is from an European-style medieval city. She’s an ethnic transplant, an orphan, but she wouldn’t be dressed in that ethnic style at this point in the story. Plus, this presentation she is too old for the scene I’m going for. Also when positioned with another image, I’d likely have to draw the top of her head.

Now here was a contender for Bryona, stepping out of a summoning circle:

Great image. Only minor corrections to change color here and there, and what is the tiny red kite umbrella? As I progressed through images, I ended up asking for her to be formed out of shadow and smoke, rather than some sort of portal.

By the time I was done, I had over a hundred images of Anuit and Bryona to choose from.

And these were only the ones I downloaded! Of these, I ended up settling on this for Anuit:

Here, she’s younger, and dressed a little more conservatively and European than some of the others. And the dark orb looks cool as a summoning accoutrement (even though that’s not exactly in the book, I’m willing to accept some ‘interpretation’ in these images). There are a few items that will need to be addressed, most notably she has two right hands.

For Bryona, I ended up selecting this one:

I liked the smoke imagery and how it looks like the smoke is forming her. Tail looks weird and isn’t joined to her body, and horns aren’t black, but the lace top is cool. The eyes don’t need adjustment, but would have preferred some green.

Ok… now to assemble. Took Anuit out of her background. I work on a Macbook, and the stanard image viewer that comes with the MacOS, Preview, has a ‘remove background’ option. So I hit the easy button on that:

And before doing any cleanup work on her, I wanted to see how she might look layered with the other image:

Not bad. I would have preferred Anuit to be looking at the succubus, but we’ll take it. I did play around with positioning, but decided to focus on cleaning up Anuit first.

Green eyes need to be brown. Fixing this is relatively trivial. You can set the Airbrush tool in ClipArt Studio to “color” mode rather than “normal” mode, and it changes the color values without adjusting the lightness or darkness.

Also needing fixing were hands:

She has two right hands here. Thumb is on the wrong side. I’m going to have to paint over that.

And I focused on removing the ethnic jewelry. She’s a simple peasant girl at this point in the story.

You see these cut portions of the picture because I rarely adjust the original image. I usually make a copy of the part I need to modify on a layer above, and then focus on that specifically. In the case of all the above, I used the color picker to select the color of surrounding areas, then the oil paintbrush tool to paint over the thing I want removed. I ended with a blending tool to take away the seams. And for the hair I usually redraw strands over the new work. Here’s the final cleaned up image of her:

I made the hands better (not perfect… not a professional here which is why I’m using this tool in the first place), eye color is changed, all jewelry gone including from her neck, and I took out the bra strap on her shoulder using the oil brush tool and selecting the surrounding skin tones. I’ve also done some preliminary smoothing of the edges. I know I’ll have to do some more blending at the edges with the layer below, but I’ll do that as a final cleanup set.

Bryona didn’t have as much to do, other than recoloring her horns to black and making them shinier, and joining the tail to the body. I worked on those details, and then slid her image to the right and put a color-matched “off-black” background in the left-behind space behind Anuit. You can see the changes here in the final image.

So that’s enough for now. Until next time!

In the meantime, I have a world to tend.