Original work - txt2Img tests for multiple models

Based off an original drawing/character of mine. I used this as a testing of models to see which was/is the best at recreating the original drawing (just 100x better)... weights/cfg/steps/etc... were set uniform across all models (with the exception of the original used to dial things in with the prompts) - essentially everything was set horribly for all of them. Thus it was both fair/unfair to every model. These are some of the better results.

Honestly the most interesting aspect was just seeing how often the same face would turn up based on the model's training or merge. You could essentially group them all together based on that alone in the end.

I graded each model then on context from the prompt, accuracy, detail, color, clarity, was there mutation, did I like the final outcome, etc...