I don’t want to use AI for my book cover

AI is excellent for helping illustrate and broaden the concepts I’m writing about. However, it fails when it comes to nuance and people of…

Diliny M. De Alwis

03 Jun 2024 — 5 min read

Two Yaksha sparring - generated by AI (Adobe Firefly)

Author Life

Why I can’t use AI for my book cover

AI is excellent for helping illustrate and broaden the concepts I’m writing about. However, it fails when it comes to nuance and people of colour.

Nuance

A subtle difference in or shade of meaning, expression, or sound

Check out the work of the best cover artist of all time: Jody Lee. Her artwork is why I picked up Mercedes Lackey and Katherine Kerr's works.

Description of coverart fed into AI-generator: “A handsome dark haired man reaches his right hand up to catch something. He is sitting astride a white horse with blue eyes. The man’s hair is loose, and his clothing is white and torn.”

AI generated image of Vanyel and Yfandes from Mercedes Lackey’s MAGIC’S PRICE (The generator ignored the prompt for ripped clothing. He is also not sitting astride the horse.) Credit Adobe Firefly

Every inch of the cover of MAGIC’S PRICE by Mercedes Lackey features relevant details pertinent to the story. When I wonder what something looks like, I have only to return to the cover to see what the artist and author had concluded met the visual. The characters jump out on the page.

Now, take an image generated by AI as an example. The characters make me curious, but they do not align with what I have written.

A lion warrior faces off with a water buffalo with long horns. A gurunda flies above them. Three lions are ready to pounce in the background. Image generated by LivingWriter’s Cover AI feature. (Living Writer)

This is a recent (possibly beta) feature of LivingWriter. You punch in some text, and it generates an AI-generated image. One of my early attempts showed a scene with various mythological characters all sporting horns. Why? Because I asked for the water buffalo to have long horns. Lions bring on the savanna as a setting.

While these generated images are pretty fantastic, they lack soul and purpose. I could spend as much time writing my novel as I would editing the AI request to develop a nuanced artwork. The AI-generated image helps me with my description and realistically judge size, shape, relation, and possibly colour. ' I use some of these images as prompts.

Having an artist generate the cover artwork for a book is worth every penny.

Here is another example.

Two female children (blonde and brunette) are crying into a lion’s main. (The lion, the witch, and the wardrobe) (Living Writer)

One version gave me two little African American girls in sweaters hugging the lion. While I appreciated the AI’s attempt at diversity, I had to ask for brunette and blonde to get this image.

AI-generated People of Color

As the Celtic French side of my family would say: “How exotic!”

I was trying to get a decent cover for my children’s story, THE EYES OF ESS-VAHA, which features an Asian Yaksha trolling a young boy who lives in the suburbs of Toronto.

Asking for a ‘brown’ or ‘Indian’ boy first gave me this. My kid’s boomer grandma doesn’t wear a sari. This is an immigrant story. TBH, I don’t know very many grandmas in Sri Lanka or India who wear a sari at home these days. Also, why is she frying onions on the ground? Where’s the stove?

Brown boy sees a yaksha floating as his grandma chops onions in the background. (Living Writer)

I had to adjust my request to include the setting. Here are the results of my attempts.

Mood board for EYES OF ESS-VAHA by DM De Alwis. Most of Asia is covered: Clockwise from the top left, Indian/Bangladeshi with Chinese influence, Korean/Japanese/Vietnam, and Malaysian/Indonesian/Thai. (Images generated via Living Writer)

Let’s take a deeper look at these pictures. For the most part, it got the little boy right. You’ll notice that asking for a brown boy brings me to a tropical setting. When I finally ask for a modern setting, it gives me a dragonette, not a yaksha. When I asked for a yellow bus, my little boy became East Asian. I had to crop the school buses because the AI-generated people were of all ages.

In some cases, the AI-generated an image that had a very literal meaning:

AI-generated image; I don’t know what’s happening here—my attempt to get a modern setting. I think I asked for the Greater Toronto Area. Maybe I was expecting the CN Tower to appear. I saved it because it was hilarious. Also, it finally got the grandma right — in modern attire. (Images generated via Living Writer)

What has all of this taught me? I would pay a good amount of money to an artist to illustrate my work properly in discussion and with the correct prompts.