This LORA trained by using several synthetic images generated from Grok which originated from one single image source and then made a several variants as datasets. Works well with most of ANIMA models. However, since ANIMA LORA training tends to being "overcooked" so you will need to include prompts with weights eg. (prompts:2) to force the image generator to produce exactly as the prompt written. Also, you will need to include prompt with weights on negative prompt section to get rid of unnecessary image result which came from its LORA training. This LORA also replaces the existing character that is built-in into the ANIMA models. See the image samples here for prompt examples.

Files with SHA256: b309be18262903bf03d911a1a3a7bbc4a2ecad11369d8cff270f93a07d27bff7