<ul><li>Harhmix is a merge model which focus on anime style painting. it could also generate NSFW content.</li><li>Features and improvements of Version 4 (My individual opinion):<ul><li>Improvement of shade</li><li>Improvement of background</li><li>Better or worse hands</li></ul></li><li>How to use:<ul><li>Clip: skip 2 or 1, In general, skip 1 is good at composition, but skip 2 is more creative.</li><li>VAE: Anything V3 VAE works very well, but other VAEs are triable.</li><li>Tags: Danbooru tag.</li><li>Negative prompts: Version 4 have no general negative prompts, now I use "(worst quality, low quality:1.4), bad eyebrow, fused eyebrow, bad teeth, poor drawn face" and change it depending on situations.</li><li>CFG: 7~11 is what I used in mostly time, lower CFG prone to Anything style, higher CFG prone to pastel-mix.</li><li>Sampler: DPM++ SDE Karras, Euler A perform better.</li><li>Highres fix: I perfer 20 steps and 0.45~0.65 Denoising strength.<ul><li>Latent: Best performace but unstable.</li><li>R-ESRGAN 4x+ Anime6B: Stable but nomal performace</li></ul></li></ul></li><li>List of merge source(Version 4): I only show the main merge source because of too many models and lora.<ul><li>Anything-V3</li><li>Nutmegmix</li><li>AbyssOrangeMix2_hard</li><li>MIX-Pro-V3</li><li>Grapefruit</li><li>Novelai</li><li>Gape60</li><li>Sun and Shadow</li><li>Better light</li><li>Howls Moving Castle , Interior / Scenery LoRA ( Ghibli Style ) v3</li></ul></li></ul>

Files with SHA256: c5cc19a640169e901fbb03c6d98b218bb023974cd47f9b31203fb00739dae68f