Hello again AI Chads! Finally I have something to show. I tried two different approaches. At first I tried as short as possible prompt. Almost all errors were gone, but the style I was after disappeared aswell no matter what I tried. So I decided to try the previous prompt again, separated it (but still optimized it a bit) with BREAK into less than 75 tokens sections (thanks again for the Diffusion tokenizer!) and it worked! Turned out some "redundant" tags I removed affected the style (like HDR tag changes style in my favorite Sugar Moonlight Lora dramatically for some reason). Cleaning the negative prompt helped A LOT too. Removing all the "masterpiece" and "best quality" tags didn't change anything (thanks for the explanation). The "quality" Lora I mentioned before. I didn't notice much different in quality but again, it affected the overall style instead. Not dramatically but it was a nice touch so I decided to keep it. Anyway, this pic is way less cartoonish than I wanted and actually had, I messed up somewhere and I think I know where exactly, but still such a good result. Without abysmal errors. Thank you once again for all your help!