Your cart is currently empty!
A 75-page AI-generated graphic novel, Cold Light is an experiment exploring how much AI can really do, and how much human guidance it required to make something.
It took 4 days to complete Cold Light, from initial concept through finished product An AI generated the storyline, characters, and artwork. A human (me) did the page layouts (adding the speech bubbles and panels). I also edited and guided the storyline and dialogue to make sure it made sense. ChatGPT kept forgetting what it had previously written.
Prologue
Chapter 1
Chapter 2
Chapter 3
Chapter 4
Chapter 5
Chapter 6
Chapter 7
Chapter 8
Epilogue
About the Story
[spoilers ahead!]
I chose from plot points that ChatGPT suggested for the story, and reminded it of those points when it forgot. ChatGPT created all the characters and named them, and I kept track of these since it kept forgetting those too.
I did suggest several plot points, like the novel’s ending, and having a series of characters that help Echo.
Not included in the story were some of the more interesting points. I liked the idea of taking down a corporation, and that being a multi-pronged approach with many different tactics. Echo being out for revenge instead of wanting to do good was also interesting. And the idea of the rich part of the city being the well-architected bottom levels, with patched together infrastructure higher up where people struggle to survive in the harsh desert climate.
However, none of these points came across that clearly, except for Echo’s revenge which felt almost too heavy-handed. Overall, the story for this AI-generated graphic novel felt a little too generic.
About the Art
[more spoilers ahead!]
I hoped to have a more desert-themed cyberpunk city, where wind, sun and lack of water affect how people live: The indoor lower levels of the city having climate control, while the upper levels were exposed to the elements. Unfortunately, Firefly wouldn’t generate anything other than generic cyberpunk cityscapes while giving images of what I needed.
I often had ChatGPT generate prompts to give to Firefly. Usually they didn’t result in anything that made sense with the story, so I had to heavily edit them. I used the same prompt to generate each character (for example, ‘cyborg hacker with green eyes’) to get panel art that was roughly similar to what the character was supposed to look like.
I also comped together multiple AI generated images. For instance, generating the background, then generating the character and removing the background of that image, and putting the character over the background.
ChatGPT had the idea of having each character have a different color associated with them. This was useful in getting generated images that looked similar enough that you could tell that this was supposed to be the same person. I used prompts for what I wanted with something like ‘style: cyberpunk, retrowave, color: red, black, gray’ on the end. This worked pretty well.
About the Layouts
I had to do all the layouts myself, and I used Adobe Express because it was easy and I could generate images using Firefly right in the layout. Initially, I tried to get Express to make some comic page layouts for me but it didn’t do a good job at all. So I ended up doing all the layouts with the panels and speech bubbles. I also had to edit most of ChatGPT’s dialogue to fit into comic book bubbles since it would give me like a paragraph of text whenever someone spoke.
I AI generated the cover, and comped together chapter pages with existing fonts over AI art.
Thoughts
Overall, it was fun to put together this AI graphic novel. It was interesting to see where AI excelled and where its limitations were. And I couldn’t have written a graphic novel in anywhere near 4 days on my own. I think AI is best used for concepting and trying out ideas rapidly to decide what you want to spend time refining and improving. AI is very generic because it pulls from things that already exist and finds patterns in them. This can be useful in many contexts, but in art you often want to break the mold.
As noted, I (and AI) created this graphic novel in 4 days. This is far less time than it would take to write and hand draw a project like this. But I think the finished product also lacks finesse, and feels quite generic and static. The characters don’t always look the same, because I couldn’t tell the AI to make a picture of a specific character that it just invented. Overall, AI generated work feels more like a way to rapidly generate storyboards and concepts which can then be refined by humans, than a way to get a full finished product.
I think the responsibilities of artists will change with AI, but I don’t think there will be less demand for art or no more jobs for artists. Artists might become more like art directors for AIs, or they might use AIs to generate an initial idea and then illustrate or model something with that idea as a direction. I think this will probably lead to more demand for artists in the long run.
AI technology presents a lot of potential, and could be very useful in many cases, like combing through tons of data in days to weeks, instead of the years it would take humans alone. But it also presents a threat to many jobs and thus human livelihoods. Personally I don’t think we will significantly slow the rise of AI – there are too many applications for it that people are willing to accept or ignore the downsides in order to use it.
Instead of fighting against AI entirely, I think humans should focus on trying to identify where AI will be harmful and mitigate that harm. AIs are training on unlicensed ‘stolen’ art work? Let’s try to get some payment to the artists for who made that work. Maybe artists can start having their own AIs to train on their own artwork, so they can try out generating images and experiment with how AI can work for them. AIs are taking jobs from humans by automating tasks that they used to do? People should have access to alternative jobs, job training, and financial support if their job is automated.
Automation has destroyed jobs in the past, and we think nothing of it now. Many new jobs were created implementing automation, designing automation, etc, and society adapted. I think the important thing is to make sure people are given support in adapting, not simply left behind to fend for themselves.
Also, I really want to live to see the robot uprising. I’m sorry. I just want to see what happens.
If you found this interested, consider following us on Pinterest or signing up for email updates!