>>2636
Pony-diffusion-v6-xl (often abbreviated "Pony" is a Stable Diffusion base model that has become one of the, if not the most used consumer level model since it was released in january. The majority of current base models and Loras have been trained/finetuned on/for Pony. AFAIK the creator initially just wanted to create an SD-model for some "my little pony" enthusiasts and furry enjoyers but the model was also trained on "normal" data (anime/hentai/cartoon etc.) and it was so superior to "normal" StableDiffusionXL (and it's derivatives) -the back then most developed version of stable diffusion-, that it quickly became the basis for a huge amount (if not the vast majority) of community driven developments. (https://civitai.com/models/257749/pony-diffusion-v6-xl)

The SD1.4/SD1.5 -models that first came out in 2022 and all the stuff that derived from it was trained on a resolution of 512x512 pixels. Then came SD2.0 which was basicly dead on arrival, mostly due to a censorship related mistake when it was trained by StabilityAI (The developer company behind baseline Stable Diffusion). Though it was supposed to be better than SD1.5 because it had been trained on a resolution of 768x768 pixels on a larger dataset, its output was shittier, because a nsfw filter slider was set to a wrong value during training, thus fucking up the quality of everything anatomy- related. Then StabilityAI made SDXL with a base resolution of 1024x1024 but since we live in the age of enshittification its quality again was in many ways lower than expected. Again greatly in part thanks to censorship. Then a guy named PurpleSmartAi trained pony on the SDXL basis and now that is what most people use. Just a few weeks ago the greatly anticipated SD3.0 was released but as of now it seems very likely to be a case of "dead on arrival" again, so for the foreseeable future it looks like pony will be around a little while longer.