In my first articles and videos on using AI art for book covers, I pointed out some of the real limitations of the current AI, which many laymen are going to miss:
- They are bad at composition, that is, mixing forms, scenery, etc. at different planes in one image
- They are bad at certain details, specifically hands, eyes, and noses.
So, how can we overcome these issues?
Well, if you are willing to do a little extra work, it’s easy. Instead of trying to fit every element you want in a picture into the AI prompt, you can make several images with several prompts and then blend them together in image editing software such as photoshop. You can also blend AI elements with real photos. I give a few examples below (and will link the live streams where I design them further down).
In the first example, there are two prompts, one for the scenery, and one for the cyborg girl. The first was “space ships over a blue alien planet” or something similar, and the second was “beautiful girl with cybornetic implants and yellow eyes.” Both had a 2:3 aspect ratio, which makes the AI compose the image with the correct proportions for a book cover.
From there, I cut out the woman and arranged half of her face off the painting, then added a lens flare by her ear for a little something more. In this case, small flaws in the nose and eyes play into the overall composition of the woman being slightly inhuman. Putting half her face off the page makes sure asymmetries are hidden.
The rest of the design is in the type (sans serif fonts with a slick modern look) and the edge darkening, or vignette. Notice we have two colors: blue and yellow, and we have all the elements of a good cover: genre signals, story elements, and a humanizing element (a face). It was simple, and the whole design was achieved in less than an hour of work.
The Next design uses an AI background and a stock photo. The background was a prompt like “verdant green, green and gold, photorealistic.” Getting a woman to be in the frame was beyond the AI’s abilities, so I just generated the background, used the “remaster” function on midjourney, and then found a photo of a woman that work with the overall design.
in order to make the two gel, I had to alter the colors of the girl to make her fit with the green/gold color. She was originally wearing white, so I yellowed her up. I also created a layer with duplicates of the flowers to cover her feet. After that, I used a splatter brush to draw yellow motes all around her, further blending the image.
Like other covers, the fonts do a lot of work. The slight grunge, yet archaic font (again in yellow), tells the reader that this is a fantasy book, possibly in the YA space, probably marketable towards women or girls.
The next design is for one of my books, coming in November. This is made up of two images, the screen and the room with the lights, then the image on the screen, each made with separate prompts. This is significantly easier than trying to get an AI to do both image elements on its own.
After that, the main things I did were color correction and an effect (suggested by a viewer on youtube) that made the TV image look blurred and out of alignment. I did this by duplicating the image and making one layer nearly transparent. Since this is literary fiction, a humanizing element is not necessary. It’s more about the mood, which should signal the stories contained within.
After that, it’s text effects. A simple set of sans-serif fonts with an outer glow (creating a nice blur effect), then a serif font for the author’s name.
My favorite of the bunch is the scifi cover. It almost makes me want to write a story to use it! But, since it’s a demo, I don’t have one ready to go. If you are reading this and want to use one of the first two covers for your story, send me an email at email@example.com. I love to facilitate others’ success.
Also, check out my manual on creative processes: