V4:This update has 2300 loras merged in total. That's 500 more than V3. I didn't smooth this last one quite as much which has led to a little bit more bias towards photo/hyper realism. I decided not to smooth for styles in this one as I really enjoyed seeing the difference and added detail in some of the outputs.The model seems a little more sensitive now but also more powerful and random from my early tests. Unless you're trying anime styling, it's best to steer clear from booru tags like 1girl, or score_whatever. Also, the model seems to give the most attention to the beginning tags so you can use them for styles, although it is more difficult to get illustrations in this one. It seems like it's more versatile in other ways now however. Some CFG scales have changed for certain subjects and I highly recommend using tools like AYS, FreeU, and PAG to really get the most out of your outputs.Enjoy!