This is a very capable model with decent species knowledge, a wide variety of artstyles, as well as pretty good knowledge of generations of artstyles by decade going back several decades and it doesn't rely on negative tags to suppress features like human skin and furry elements if you want to make a human character or if you want to make a furry character.Some of my favorite positive tags for style has been "Anime Screenshot, Anime coloring, Digital artwork" which gets you a decent stylized anime image, if you want less 2d drop the anime coloring tag and or the anime screenshot for more of a 2d cartoon look. and if you want more 3d stuff you can tag for that too.You can use the default set of "masterpiece, best quality, absurdres" etc and "worst quality, low quality, low res" but I don't really think its necessary for this model.PLEASE NOTE! This is only tested in SDNext, aside from briefly loading it up in forge to verify something I've only ever used it in SDNext, if you have a hard time reproducing my images using your software of choice I cannot help you there. Maybe consider switching software?My example prompt:Positive: Cute female (fluffy:0.7), furry, Raccoon, detailed eyes, hair over one eye, Lace-up panties, looking away, newest, very aesthetic, best quality, absurdres, masterpiece, detailed face, A treacherous mountain peak shrouded in swirling clouds revealing glimpses of a forgotten temple carved into the rock face.Negative: blurry, poorly drawn, out of focus, (worst quality, low quality:1.2), simple background, 3d, bad hands, bad anatomy, (signature, watermark), sketch, drawing, extra digitsSteps: 16-30Sampler: DDIM, Euler, Euler A, DPM++ 2m, DPM++ 2m SDECFG: 4-10Resolution: 832-1536, This version does much better on the lower end while still maintaining strong high resolution results. I prefer either 1024x1024 or 1024x1536clip skip set to 1 (again, this is for SDNext, you probably need a different value if you're in A1111/forge or comfy)

IcedCoffeeIL

New version! I'll update the model details to include my findings but overall this model improves significantly on v10, better background consistency, less deformations for hands and feet, better anatomy across the board. Overall this is my preferred version of IcedCoffeeIL.

Files with SHA256: 8d20278103bb686f3a2fa86143c49375b2385cd1d8c4890ea295cb26084cd31f