My Book Cover Cost More Than AI Help Could Ever Convey
The previous post covered the release work on KDP. There I mentioned lightly that “AI helped with the cover,” but this cover cost me far more than “helped” can convey, and copyright was something I thought about seriously too. This time, that’s the story.
What kind of cover I wanted
What I first pictured was a cover that wasn’t cluttered.
When you have AI generate images, you end up wanting to pile on elements. But at the start, I wanted the opposite — I wanted negative space and subtraction to do the work. What I had in mind was a simple, abstract composition: an arc like a contrail stretching across the sky, and a geometric circle softly fading inward.
(Let me confess up front: this “initial ideal” would get completely overturned at the very end.)
The process went like this:
- Have AI image generation (Nano Banana Pro) create the base image
- Bring it into Canva to crop, arrange, and adjust the coloring, then layer thin-to-medium-weight serif typography on top to finish it
That was the flow. Writing prompts, I found that specifying material and texture — “lithograph-style,” “monotype,” “airbrush gradient” — got much closer to the image I was aiming for than shape instructions like “circle in the center, line on a diagonal.” Material and restraint and negative space, more than diagrammatic precision, is what worked this time. And this process — have AI make the base, then finish 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 an answer
This wasn’t the composition from the very beginning. If anything, I wandered quite a bit.
The first thing I ran into was a question of strategy. When I consulted AI, it told me that for “Naro”-style covers, character art pushed front and center is basically the default; abstract design works better for general literary fiction, but in the Naro context it’s basically a disadvantage. I see, I thought. So I resolved to run an abstract draft and a character draft in parallel.
The abstract draft went through stages like this:
- The first “pitch lines × monochrome × geometric” draft → too polished, rejected as “looks like a slide from a presentation deck”
- Trying to add more work-like character, moving toward risograph, lithograph, monotype, ink-wash gradation
- A draft with a “doubled ghost figure” hinting at “another life = reincarnation” → rejected as too diagrammatic
- Following the design of a book I used as reference (a single bold diagonal line on a white ground), moving toward a draft that let a single brushstroke say everything
- But that ended up too Eastern, too Japanese in feel. Mixing in Bauhaus, Swiss design, airbrush gradients pulled it back toward Western abstraction…
- Finally arriving at an abstract draft I liked myself: “an arc like a contrail, plus a soft circle”
The character draft had its own separate struggles. The composition was “James seen from behind, looking out over an empty pitch from beside the dugout, an England stadium from the 1980s.” I made it first in a muted, film-grain realism, but it came out too melancholic, too heavy. From there I redid it repeatedly, moving toward cel-shaded anime style and adjusting the downcast pose toward the bearing of a “quiet hero.”
The most quietly troublesome thing was that the generation tool kept inserting an unremovable watermark in the bottom right. Since the prompt couldn’t get rid of it, I had to resort to tricks like deliberately not placing important elements in that corner from the start.
One conviction ran through all of it: I didn’t want an image that read as “obviously AI-made” at a glance. Rather than aiming for perfection in one shot, I generated many and chose among them. I didn’t want to make the face of a book look cheap, and that tension was what ate up the most time.
In the end, one advertisement I saw on the street flipped everything over
The abstract draft wasn’t bad. I liked it myself. Or so I thought.
Then one day, I happened to see an advertisement for a Naro-style work on the street. Every single work lined up there had its character pushed forward, front and center, huge. In that moment, it clicked into place. “Ah, right — in this genre, you really do have to put the character front and center.”
I’d understood in my head what the AI had told me at the start — that character art up front is the default for Naro — but somewhere in my heart I’d been prioritizing my own aesthetic sense, wanting to nail it 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.” The reality of that street advertisement drove that home.
So I dropped the abstract draft outright and switched entirely to manga-style with the character front and center. What resulted is the current cover: the protagonist James, sitting on a bench, gazing out at a pitch after rain. Behind him, an England stadium from the 1980s; beside him, the other people on the bench; at his feet, a tactics board. The title stands large and bold: ‘EXTRA TIME’. After all that wandering through abstraction, the “character draft” I’d started with ended up seizing the leading role at the very end. Ha.
(This is the cover-art version of the “riding the Naro template” story I wrote about last time. Here too, in the end, the right answer was simply to follow the genre’s conventions.)
Since the cover uses an AI-generated image, this was a subject 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 thought through myself.
- How copyright applies to AI-generated output: whether copyright arises in the generated work itself, and if so, who it belongs to. As with the body text, my understanding is that the degree of human creative involvement is what matters here (the framing being that an image lacking human creative contribution is unlikely to be protected).
- Risk of resembling existing work: whether, due to its training data, the image ends up resembling a specific artist, work, or character. This work in particular is set in 1980s English football and features real club names in the story, so I was warned that if the cover ended up drawing on a real club’s emblem, colors, or a real player’s likeness, that would be dangerous. So even after deciding to put the character front and center, what I drew was strictly my own original character (James) and a fictional club (Wandle). I was strongly conscious of not drawing close to any real club or player, and of making the emblem fictional too.
- The terms of the tools used, and whether commercial use is permitted: whether the generation tool I used, along with Canva and the fonts, carried licenses that permit use in commercial publishing. I checked this including the license on the serif font. As the cleanest option rights-wise, I also considered something like Adobe Firefly, whose training data is licensed and comes with indemnification.
- KDP/Amazon’s own terms: the platform’s content requirements. KDP draws a line where “AI-generated” content requires disclosure, while “AI-assisted” content does not, so I checked which category my process fell under.
- How to disclose AI use: how far, and in what way, to make this explicit, including for the body text.
In the end, the design swung hard from abstract to character-front-and-center. Even so, my basic policy never changed: use AI to generate “material,” but make the final compositional decisions — cropping, placement, typography — myself.
To put it concretely — the base image is made by Nano Banana Pro. But it doesn’t end there. I bring it into Canva and, working it over repeatedly myself, decide what to crop, how to place the text, how to adjust the colors, finishing it into a single “cover.” I concluded that it’s this editing process where human creative judgment actually sits. Using the generated output as-is, versus reconstructing it yourself using it as raw material — these differ enormously in the degree of “human creative contribution” — and this, I think, is what matters both for thinking about copyright and for thinking through KDP’s line between “AI-generated” and “AI-assisted.”
As it happens, this is exactly the same philosophy as how I made the body text: have AI write, but keep the final judgment in human hands. For both the cover and the text, my own guiding axis was how much human creative contribution I could preserve.
Just to be clear — I’m not a lawyer, so what’s written here isn’t a legal conclusion. This is an area where the rules are still in motion, so I’d like this to be read as a record of “here’s how I looked into it, and here’s what I decided,” something that might be useful as a reference for others about to publish (please check the current terms and licenses yourself as well).
Next time, I’ll come back to something more substantive. I’ll write about the one “heavy theme” I packed into this story that I started in such a lighthearted way.
Below are the remains of the failures
Originally published in Japanese at https://clazytech.com/2026/06/1654/. Translated with LLM assistance and reviewed before publication.