Day 26 of 30 Day AI Challenge – Creating A Comic Strip With ChatGPT and Midjourney

For those just joining us, every day during this challenge, I’m going to try and do something different or better, using currently available Artificial Intelligence tools.

I’ve always wanted to create comics. Sadly, with little to no drawing talent, this aspiration has been out of reach. For today’s Generative AI challenge, I’m going to take on just this ambition. Today’s experiment will feature a series of comic panels, set in “The Hotel”, our fictional setting from Day 23’s challenge.

The Process, In Brief

The overall process was remarkably similar to other efforts.

First, train ChatGPT on the task and then start filling in the blanks together.

The iterative process continues as a dialogue with ChatGPT, where I provide additional details on comic style, settings, etc. Ultimately, we ended up with a few ideas for a single-panel comic.

Second, work with Midjourney to create the desired visuals.

I tried a couple of routes here. One was to train ChatGPT to create the right prompts…this went back and forth a few times till I was satisfied with output I could refine and feel in myself with the right parameters.

Anyone who’s worked with Midjourney will know that text is best avoided. It is also pretty hard for the bot to understand specific situations and character placements, especially with multiple objects…it’s great with overall stylistic treatment and simply described scenes though.

Third, combine the visuals with the right captions and punch lines. For this step I used the free and easy Canva, which is the simplest way to put the desired single panel comic together and export it into a usable format.

The Final Output

Here are a few selected comic panels created using this process:

Wrap Up

I was pretty satisfied with the final creations above.

What I loved about the whole process was the iterative co-creation with ChatGPT. The generative AI is tremendously capable as long as you make the effort to train and guide it, plus apply final refinements yourself.

What I hated about the whole process is the ability to create consistent visual characters for the comic, despite locking these down during the ideation process. As the image creation process at the moment is pretty tough to channel with consistent elements (vs styles), the best approach here would probably require working with more standard creative tools.

I suspect as the tech and plug-ins evolve, though, we’ll see a lot more ability to create exactly what you want and retain consistency throughout a series.

That’s day 26 wrapped…see you tomorrow!