Welcome to Levy is out now!

This is a short horror VN we decided to make after learning how to use AI tools to make art. Since we've had some questions about it below is a bit about the process we went through to get this far. If you have any questions about our process please feel free to reach out!

Once we were familiar with using Stable Diffusion at a basic level we decided to learn more about it by making a game with it.

The first stage was dialing in some prompt and setting combinations that generate things that have a cohesive art style. Once we could generate things that look like they belong in the same setting, we started making large batches of images with only minor tweaks.

For this project we ran one more batch that was much larger, and then pruned it back and sorted things into delete, good enough, and special images we would work on more or use for more central parts of gameplay. Then we went through the special images and made whatever changes by hand they needed and ran them back through the img2img pipeline to smooth out our edits. This could be done for all of the images, but for this project we were focused more on learning the tools and processes so we opted to leave most images untouched.

The whole theme and story of the game was based around the art that was generated. Since it was our first time we really needed a forgiving setting, and body horror is something that AI is really good at - even when you don't want it to be. The areas the player goes through are based mostly on the background images we chose. We just sort of made up a little back story for each one and connected them all with a plot. We could probably pull of something cleaner and less obviously AI generated now, but this was all a learning experience.

So what do you think? When you saw the art did you think it was made by an AI right away? Has this changed your perception of AI art at all?