So this is the cover idea I shared when announcing I had finished my draft. Well shortly after finishing my edit to refine that draft to the point I am ready for other people’s eyes to take it in, I was caught once more by how, while this is fine… none of the characters are quite right. Nightcafe, my current preferred website for creating AI art pieces, had recently launched SDXL 0.9 so new technology meant I might be able to get an improved result. I will disclose now that I am more pleased with my current result but it still isn’t perfect.

However let’s look at some of the steps along the way. This is labeled as iteration 2.0 prior to adding the cover text. Every time I make a major change it gets labeled as a new front number and variations there in are the decimal points, much like software. We’ll start with 3.0 where I first ran it through the AI again and got a new version of Cressa that I love. From there, I tried to start working on improving Jenna, the character on the right of the screen. Jenna is Hispanic, specifically with heritage from Mexico however in this moment in the book she has tamed her curly hair into a kind of mohawk. Curly mohawk tends to skew my results more Afro-textured, I would assume that would be because there are more examples of a curly mohawk being done on that texture. I was also worked on trying to darken Esther’s hair on the left.

The above four images are 3.0, 3.1, 3.2, 3.3, and 3.4. I got a new face for Jenna and quickly decided that person was too old for being twenty-eight. I also loved the update to the food at their table. We went from cartoon food to realistic food. I worked to blend a new face with hair I was pretty pleased with. In later iterations, I did some repair on the color of Jenna’s top as it was starting to blend into her skin.

We move into iteration 4.0. This time didn’t start with what I got out of the AI but rather what I was putting in. This required images from another program I used. Artbreeder uses AI to blend two images together. It also allows for manual adjustments. I used this program to create my more realistic character portraits.

These are the two images for each character that meld the most sway on my final images for each character.

For version 4.0, I used these images to try and fix the characters in the group scene.

Above are featured 4.0, 4.1, and 4.2. Between 4.0 and 4.1, I went back to Nightcafe and got new figures. They created new versions of Cressa as well but I didn’t keep those. You can check my creations on Nightcafe directly if you’re interested in seeing that craziness: [click here]. Between 4.1 and 4.2 I was trying to get Esther’s hair back to the level of curl it has in my other picture created for this book. I’m thinking another article on that shortly.

In the 5.0, 5.1, and 5.2, I found that I really liked image of Esther from 4.1 so we went back to fussing with Jenna’s hair. I ultimately decided that 5.1 was probably the best I had and proceeded with that to my cover. So, here is my update cover a project encompassing an additional 30 hours of work.

Truthfully, I’m still not satisfied and am chasing better. From here it most likely mean reaching out to a professional but we’ll see. New AI tech keeps coming out. I’ll finish up with a side by side on the two iterations.