I’ve started to now use the new AI tools to create art for my Attack of the BiiBiis project. My first attempt is for my witch hero Lodet De Vil. I first did a sketch of her in the hopes of being able to use AI to import that and create art from it. Here is my sketch with an art style I’d like to go for, a modest design that is family-friendly and with some of the magic elements to mix witch with magician.
After some poking around, I picked Leonardo AI to try this out. So far, my sketch idea is a total failure. AI thinks I’ve created a nude image and won’t even render it. I suppose I’d have to fully colorize this to let AI know what I’ve intended.
Instead, I’ve gone with the “prompt” model where you describe what you want and generate some images and try to refine. Here is where I landed so far – an image and a video:
Lodet’s face, hat, hair and overall figure is spot on. Far too sexy for what I envisioned though, and that’s true for the half dozen other variants I created. Apparently when I say “beautiful” this is what AI thinks that means. And sure, not completely wrong, but also not exactly matching my thoughts.
The video is humorous as my prompt was that she is raising her hands to cast a hex spell. That’s not what I got. She is very cute though. I’m gonna have to work on adding some modesty to the model. Maybe playing around more with the videos and going for more of the magician element could get me closer to the target.
The core challenge is that if AI can be used, my project can really jump ahead and become something at a speed far beyond anything I could otherwise afford. So I will keep digging into the AI system to see if I can meet that goal.
I took this model and then asked the AI to make a version of her angry casting a green fireball. These results indicate this is going to be challenging to get right. It zoomed in (didn’t ask for that) and no fireball at all, and all variations are nearly identical with different finger movements.
