Clay Tech

"clay-works make things real"

translated from clazytech.com

I Struggled Through Dozens of AI Cover Drafts Before Finding the Answer

Last time I talked about the release process on KDP. I mentioned in passing that “AI helped me with the cover” — but that cover involved a lot more struggle than that offhand remark suggests, and I thought seriously about copyright too. This time, that’s the story.


What kind of cover I wanted

What I pictured at first was a cover that wasn’t cluttered.

Letting AI generate images tempts you to pile on elements. But at the start I wanted the opposite — I wanted whitespace and subtraction to do the work. What I had in mind was a simple, abstract composition: an arch like a contrail stretching across the sky, and a geometric circle fading softly inward.

(Let me confess upfront: this “initial ideal” would get completely overturned right at the very end.)

The process was:

That was the flow. While writing prompts, I noticed something: specifying materials and textures — “lithograph style,” “monotype,” “airbrush gradient” — gets you much closer to the image you’re aiming for than shape instructions like “put a circle in the center, lines running diagonally.” Not diagrammatic precision, but materials, restraint, and whitespace. That was my takeaway this time. And this process — having AI generate the base, then finishing it myself in Canva — connects directly to the copyright discussion I’ll get to later.

The struggle: it took a long time to arrive at the answer

This wasn’t the composition from the very start. If anything, I wandered quite a bit.

The first thing I ran into was a strategy problem. When I consulted AI about it, the answer came back: “For Narou-style covers, putting the character front and center is basically the default. Abstract design works better for general literary fiction, but in the Narou context it’s generally a disadvantage.” I see, I thought. So I decided to commit and run an abstract concept and a character concept in parallel.

The abstract concept rolled through stages like this:

The character concept had its own separate struggles. The composition was “James seen from behind, standing by the dugout, looking out at an empty pitch, an English stadium from the 1980s.” The first version, done in a muted, photorealistic film style, came out too melancholic and heavy. From there I redid it again and again — shifting to cel-shaded anime style, adjusting the slightly downcast pose toward the bearing of a “quiet hero.”

One quietly annoying problem throughout was that the generation tool kept inserting an unremovable watermark in the bottom right corner. The prompt couldn’t get rid of it, so I had to work around it — deliberately avoiding placing important elements in that corner from the start.

Running through all of this was one conviction: I didn’t want an image that screamed “made by AI” at a glance. Rather than aiming for perfection in one shot, I generated many versions and chose among them. I didn’t want to cheapen the face of a book — that tension was what ate up the most time.

In the end, one advertisement I saw on the street flipped everything over

The abstract concept wasn’t bad. I liked it myself. …Or so I thought.

Then one day, I happened to see an advertisement for a Narou-style work out on the street. Every single work lined up there had its character front and center, taking up the whole space. In that instant, it clicked. “Ah, right — in this genre, you really do have to put the character front and center.”

I had understood in my head what the AI told me at the start — that Narou covers default to character art — but somewhere in my heart I’d been prioritizing my own aesthetic sense, wanting to go with something abstract and cool. But what matters for a cover meant to get someone to pick up the book isn’t my aesthetic sense — it’s whether the reader can tell at a glance, “this is the kind of thing I like.” That advertisement on the street confronted me with that reality.

So I dropped the abstract concept entirely and switched completely to a manga-style cover with the character front and center. What came out of that is the current cover — the protagonist James, sitting on a bench, staring out at a pitch after the rain. Behind him, a 1980s stadium; beside him, his teammates on the bench; at his feet, a tactics board. The title sits large and bold: “EXTRA TIME.” After all that wandering through abstraction, the “character concept” I started with ended up stealing the lead role at the very end. Ha.

(This is the cover version of the “I went along with the Narou template” story I wrote about last time. Here too, following the genre’s conventions turned out to be the right call in the end.)

Since I was using AI-generated images for the cover, this was something I couldn’t avoid. I’m not an expert, so I can’t state anything definitively, but I’ll leave a record of the points I checked and considered on my own.

In the end, the design swung dramatically from abstract to character-front-and-center. Even so, my basic policy never changed: use AI to generate “material,” but keep the final compositional decisions — cropping, placement, typography — in my own hands.

To put it concretely — I had Nano Banana Pro produce the base image, but I didn’t stop there. I brought it into Canva and kept making adjustments myself — what to crop, how to place the text, how to balance the colors — until it came together as a finished “cover.” I believe human creative judgment comes into play precisely in this editing process. Using the generated output as-is, versus using it as raw material and recomposing it yourself — these are entirely different in terms of the degree of “human creative contribution.” That distinction matters both for thinking about copyright and for thinking about KDP’s line between “AI-generated” and “AI-assisted.”

As it happens, this is exactly the same philosophy as how I made the main text — letting AI write, while keeping the final judgment in human hands. For both the cover and the text, my guiding principle was how much human creative contribution I could preserve.

Just to be clear — I’m not a lawyer, so what I’ve written here isn’t a legal conclusion. The rules in this area are still in flux, so I hope this can be read as a record of “here’s how I looked into it, and here’s the judgment I reached” — something that might be useful as a reference for others publishing going forward. (Please check the current terms and licenses yourself as well.)


Next time, I’ll return to the actual content. I’ll write about the one “heavy theme” I packed into a story that started out on a light note.

👉 Read it on Kindle

Below are the wreckage of my failures


Originally published in Japanese at https://clazytech.com/2026/06/1654/. Translated with LLM assistance and reviewed before publication.