Will AI Artwork Destroy Comics? My Hot Take.

A quick attempt at creating Robo with MidJourney (as drawn by AI Jim Lee)

GANs & AI Artwork

The internet has been starting to buzz over the last year or so. Something exciting has been cropping up here and there in places like Twitter, Reddit, and Instagram. Artificial Intelligence programs can now create artwork with just a text prompt to get them started! This means you can finally see Harry Potter holding a shrunken down Batman or Captain America punch Hitler in the face!

But this technology has actually been around for a little while and it's something I've been very interested in almost as soon as I first heard about it. Years ago, Google released an image that came from their "Dreaming Engine", the resulting images to me were nightmarish.

I thought about this as an eventual encroachment from AI. The first steps of machines beginning to threaten creative work, which I'd long thought would be our last saving grace. The idea stuck with me and actually was an inspiration for the look of Circuit when Phil and I were designing the character for Robo #1.

Early in 2021, while writing Robo #3, I had been learning about the technology behind these sorts of images. They're called Generative Adversarial Networks and they're actually really genius ways of using lots of the AI tools that we already had to achieve this. They're also not just for images. AI can create music and poetry and probably even 3d models to turn into sculptures too.

From earlier work that I've done in Natural Language Processing technology, I learned that the most important piece of any machine learning tool is the dataset. The more high value text you feed into an NLP program, the better it handles text. For a GAN which creates artwork, the more images that you give it, the better it is at generating those kinds of images. You may have already seen this at work with something like This Person Does Not Exist, a website that generates realistic human faces every time you reload. There are a LOT of human faces on the internet.

Recently, AI artwork services have gotten a lot better. Both through improvements in the software and architecture of them, but also because there's just so much artwork on the internet to feed off of now. I've been playing around with a service called Midjourney which is largely regarded to be the best current AI artwork service.

AI Artwork and Comics (Here we go...)

Okay, so AI can create artwork. The more artwork that you feed into the AI the better it gets. And you can create the artwork just using text prompts to describe what you'd like to see. You can see where this is going, right?

The Indie Comics community came out pretty strongly against AI artwork earlier this year. There are a couple of reasons for this response and here are the ones which I think are the most valid.

The AI Artwork engine is stealing the work of artists. You need to train your AI using actual artwork so it can learn styles and common things and the better the data-set the more accurate it becomes. Comic artwork is vast and often already comes with a document that describes what it is. I think there's some valid concerns here in the short term, but they are likely legal and copyright issues that I don't feel like I can weigh in on.

The second worry is that AI Artwork will replace human artists in comics. Writers are already nearly or completely free labor in comics. Now you just get a writer who knows how to use the GAN and have the AI do the artwork! This is the one that I am going to dig into a little.

My Experiences with MidJourney So Far

Above, you will see an image of "Galactus" from Midjourney. The prompt I wrote was only "Galactus" and these are sort of very slightly like Galactus. He's a very unique looking character, he wears a helmet, lives in space, and he's purple. It's Galactus, right? Not really. And furthermore, it would be very difficult to consistenly recreate this at the skill level I have with prompts right now. GAN artwork is not deterministic either. If I put in Galactus again, it will show me different options.

The advice I've seen so far is that you have to keep refining the results over and over until you get what you want. The part of the GAN which determines the outcome is not even close to a human in its ability to interpret the prompt. BUT it's also really really good at it compared to anything we've ever seen.

You can also feed in a source image to try to get a better outcome. Here is an example I tried earlier today:

That's me on the left. Or the right? Who can tell! Okay I fudged this one a bit. The prompt here was my profile picture and then "by Jack Kirby". I've had a lot of success having the AI immitate Jack Kirby in order to get a comic-book feel to the result. It does not really look like Jack's artwork to me, but it does look like a comic book. That's Moose's inner personality, by the way. She thinks she's a very menacing and tough cat.

So now we've hit on some of the really scary parts when it comes to comics artists and comics creators and people who care about the work of creative folks being cannibalized for the masses by an AI which will replace us all. AI Artwork can create images of characters which are well known and it can immitate the styles of famous artists (it's really good at Moebius) without even crediting them. In fact, it fakes a little signature sometimes which comes out to be complete nonsense. The AI that creates artwork generally knows that there should be letters, but it doesn't really spend a lot on them making sense.

So Will This Destroy Comics?

You could, through careful work in prompts and consistent source images, create a comic book using an AI artwork generator. I think it would be quite a bit of work and you'd have to do it panel by panel with some clever photoshop work in between. I know that there are people who have already done it. I haven't been working with Midjourney very long yet, but I think this would end up being a lot of work. It would also be faster than an artist spending an entire day drawing a page. And cheaper.

Comics has a history of finding ways to exploit artists, writers, and basically everyone working in the industry. There's also a tradition of pawning off work onto cheaper artistic talent and claiming that its your own. It's scummy. But it won't destroy the comics industry.

Here's My Hot Take

I think now is the time to start exploring this. I think the best and most creative AI comics will be created in the next year or two. While the ground is fresh and people are wrestling with the quality of the GANs creating them, we could see some really amazing things come out of it.

I think it's only a matter of time before Disney sinks some money into their own version of Midjourney. It'll be trained entirely on their vast back catalog of artwork and scripts. Disney has a lot of resources and a big interest in being able to churn out comics for their characters for the lowest overall cost. People aren't going to stop buying Spider-man comics even if the hands don't look right for a few years.

If you ask me, mainstream comics will largely go towards AI art eventually. Indie comics will be the place where artists are still creating beautiful art that looks just like you want it to and uses creative angles and innovative storytelling. And Batman will have 10 monthly titles that sell great and are drawn by software.