Most of these versions are just the result of me naively smashing things together and kind of learning what works. More or less just iterating on it until the outcomes fit what I wanted. Originally my goal was to put together a merge that was very flexible and could be a sort of all-in-one model that could generate images with an abundance of detail and creativity. I attempted to get good results for both 2D and 2.5D semi-realistic images, and put a lot of emphasis on having it be able to make interesting monster girls and furries.
I may or may not be smarter now, and so from v14 onward, this model is going to be a bit more focused and finetuned on being more highly detailed and realistic. My goal is stylized realism, but not photorealistic. I will probably eventually make new separate models that are tuned to other art styles.