AI Can Help Us Color Comics

I know there's a lot of anxiety about AI, but it does have its uses.

I hate coloring comics. A big reason is that I find it tedious. It’s the least engaging part, and I’m not very good at it. When I got my advance for my first two graphic novels, the publishers agreed to give me some extra money to hire a colorist, but it was too hard to resist keeping it. Next time, I’m going to try to outsource it.

But in the meantime, I’ve been working with limited palettes to limit the choices of colors that I have available to me. I’ve been reading books and watching videos about color theory. And as a result, I created what I thought was a pretty good coloring job for a recent submission.

But just for fun, I wondered if ChatGPT could help analyze my color choices and make them better. After all, most of my coloring anxiety stems from not being very good at it and spending a lot of time picking colors when I don’t really know why I’m making those choices, and maybe AI could help me do better. I have been fairly critical of the limitations of AI in the past, but I’m always willing to give it a try if it frees me up from some of the work that I don’t enjoy in order to spend more time doing the work I do enjoy.

So I asked it for help. Here’s a sample of what it came up with.

1. Your Biggest Issue: Midtone Overload (Value Problem)

Almost everything sits in a similar mid-range value:

• Grass, sky, trees, dirt, even skin tones

• Result: the page feels a bit “flat” and your focal moments don’t pop

Fix:

Pick 3 value tiers and commit harder:

• Light (backgrounds / sky / negative space)

• Mid (environment)

• Dark (characters + focal objects)

Specific adjustments:

• Darken characters’ clothing by ~15–25% (especially the maroon and brown coats)

• Lighten the sky slightly (push it closer to a pale blue/gray)

• Deepen the grave interior significantly (this should be one of your darkest areas)

👉 The grave should feel like a visual “hole” your eye falls into—but right now it blends with the ground.

There is a lot of specific feedback here just for one panel, and it seemed like sound advice based on what I’ve read about how color works. And while some of the suggestions might end up being inconsequential, I can easily play around with them in Procreate.

But the most mind boggling thing to me was that it was able to to read the image correctly. Even in a drawing this cartoony, AI knew it was a page of a couple of guys digging up a grave. It picked up the details. I was simultaneously impressed with what AI can do and reassured (maybe flattered?) that the image was this easy to read. Perhaps there’s a good use for AI here as well—give it an image and have it explain what’s going on.

I asked it for some more clarification for each panel, such as how much to shift each color. But I don’t need to share that. You get the idea.

So I made a few adjustments based on ChatGPT’s suggestions—subtle changes, to be sure. I think it looks better. But you can be the judge. The first image is the original.

I’m still not completely happy with it—the shirt of the guy with the hat is too close to the color of the sky. Overall, though, I like the direction. And I could put it into ChatGPT again to see what happens, but I’ve learned that AI is always looking for things that are wrong, and may not give me the all clear, and at some point a human (me) has to decide when it’s done.

Here' are some other posts you might enjoy:

And if you like my work, why not consider becoming a paid subscriber? I love telling people about how I make comics, and your financial support would really mean a lot.