DESCRIPTION: In this project I wanted to see how I could introduce AI into my process without simply wholesale copying the image. The goal was to take this design and portray the character in various different mediums. While this was a fun experiment to explore, it isn’t a direction I think I’d pursue much in my future work.

PROCESS: I generated a number of characters using a generated which didn’t have much fine control, and tended to output results that were either blurry or had a number of nonsensical features, which I thought would let me have more creative control in how I interpreted and made sense of the images.

I took the generated image I thought I could do the most with and turned it into a silhouette, then refined the shapes. I tried to reference elements of the original image, while also not sticking to it too closely in order to make it mildly more of my own design. Ultimately, I didn’t end up including as many of my own ideas as I would’ve liked, so the end result is still fairly similar.

From there I rendered out the image, added details, and tried to push it even further away from its source, especially in terms of materials and color schemes. I did lean into the sort of organic, fleshy biotech aesthetic from the original in places, but I felt a mix of organic and inorganic would make for a character with better visual contrast.

At this stage I didn’t really know what more I wanted to do with the character. I’d considered doing a full character turnaround, but since this was supposed to be a quick experiment, I ended up decided I’d apply the design to different situations, the first being a poster in a more illustrative art style, and the second being a rough 3D render, which despite the limited success still allowed be to explore the character more.

While this was an interesting experiment for me, and something I’m glad I explored, I think in the future I would either avoid sticking nearly as closely to the reference material, or just avoid using AI at all. It was far too easy to fall back on the original image when I wasn’t quite sure where to proceed, and the abstract nature of the images meant I spent more time than I would’ve liked trying to make sense of how things were constructed and what certain things even were. I can imagine it being a useful tool if used correctly, but I’m personally not sure what the right way to implement it would be yet, even leaving aside the numerous issues surrounding AI generated images as it is.