Like in every post, I try to focus on new unique and interesting AI art. This time I focus on cute baroque victorian-era chibis. “Chibi” (ちび or チビ in Japanese) is a Japanese slang word describing something or someone short.
I’ve chosen the victorian era because the clothing is more serious, and conservative which contrasts the young age look of the characters in this art. I also wanted a baroque 17th-century style with cute fuzzy doll hair. combine all of that into something that looks adorable, kawaii, fascinating, and unique.
Funny thing is, I didn’t use the word “chibi” at all and the “cute” word had already a strong weight in the look of the character rendering. Along the 3:4 aspect ratio, I got very good results. When I did use the word “chibi” it rendered it too cartoony. As you can see, it’s all about experimenting. sometimes you don’t even need to use certain words and committing them can lead to desired results or surprise you in different ways. I did like the cartoony-drawing results, don’t get me wrong, but it’s not what I wanted to achieve here.
I had to do a lot of Photoshop work though. Fixing the eyes and some imperfections. Even when using “detailed upscale redo” I wasn’t 100% satisfied with the results. Even after some retouching, I stopped at some point. I could have made the very smooth but at some point, I just was ok with the results. I can always do better, but it takes a lot of time.
I did try adding keywords like “smooth”, and “symmetric face”, and it helped a bit but it affected the overall look of the art–so I dropped it.
Working on the eyes is one of the most annoying things as it happens a lot and requires fine retouching work in Photoshop. It’s not a big deal but I would prefer the AI to be better at that. You don’t care if it’s a creature or an object, but when it comes to faces, every single thing that it’s “out of place” is easily recognizable.