@Background Pony #FF57
Unfortunately, that’s what is par for the course for AI generated pics. If we had an AI trained in cutie mark recognition, and integrated it into the image generation system, with the input properly cross-referenced to character, then it would be different.
Na, just stack more layers and use better datasets. Pony Diffusion V6 XL handles Rainbow’s CM almost perfectly: https://i.imgur.com/TNF7VVF.png
Unfortunately, that’s what is par for the course for AI generated pics. If we had an AI trained in cutie mark recognition, and integrated it into the image generation system, with the input properly cross-referenced to character, then it would be different.