Huggingface Repo: <a target="_blank" rel="ugc" href="https://huggingface.co/lex-hue/Fluffity">https://huggingface.co/lex-hue/Fluffity</a> Use <a target="_blank" rel="ugc" href="https://e621.net/tags?commit=Search&amp;search%5Bhide_empty%5D=1&amp;search%5Bname_matches%5D=%2A&amp;search%5Border%5D=count">e621 tags</a> (no underscore), Artist tag are very effective (SD/e621 artist)ClipSkip: <a target="_blank" rel="ugc" href="https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#clip-skip">Stop at CLIP layers</a>=1 (SD=1, NAI=2) [Settings\Stable Diffusion\Clip skip]<h3 id="heading-71">Example Setting</h3>Txt2img Setting<ul><li>Steps = 15~35</li><li>Sampler = Any</li><li>CFG scale = 5~25</li><li>Negative embeddings = <a target="_blank" rel="ugc" href="https://civarchive.com/models/229376/embeddings-for-fluffity">fluffity-neg</a><a target="_blank" rel="ugc" href="https://civarchive.com/api/download/models/84134?type=Training%20Data"> </a>※[SD-WebUI\embeddings]</li></ul>Hires. fix<ul><li>Hires steps = 15</li><li>Denoising strength = 0.45~0.5</li><li>Hires upscaler = <a target="_blank" rel="ugc" href="https://mega.nz/folder/qZRBmaIY#nIG8KyWFcGNTuMX_XNbJ_g/folder/zJI3EY6T">4x-UltraMix_Smooth </a>※[SD-WebUI\models\ESRGAN]</li></ul>Img2img Resize<ul><li>Scale to = 1.5~2.0</li><li>Denoising strength = 0.45~0.5</li><li>ControlNet0 = softedge_hed, <a target="_blank" rel="ugc" href="https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main">control_v11p_sd15_softedge</a></li><li>ControlNet0 Weight = 0.4~0.5</li><li>ControlNet0 Pixel Perfect = true</li><li>ControlNet1 = Reference</li><li>ControlNet1 Weight = 0.2~0.3</li><li>ControlNet1 Pixel Perfect = true</li></ul>SD Setting<ul><li>Stop at CLIP layers = 1</li><li>Eta noise seed delta = 0</li></ul>Txt2img <a target="_blank" rel="ugc" href="https://github.com/continue-revolution/sd-webui-animatediff">Animated Diff</a><ul><li>Steps = 20~50</li></ul><ul><li>Sampler = Any</li><li>CFG = 7~25</li><li>ImageSize = 512~768</li><li>model = <a target="_blank" rel="ugc" href="https://huggingface.co/manshoety/AD_Stabilized_Motion/tree/main">mm-Stabilized_mid, </a><a target="_blank" rel="ugc" href="https://civarchive.com/models/139237?modelVersionId=154110">motionModel_v03-anime</a></li><li>Number of frames = 8~16</li></ul>

Fluffity (SD1.5)

<ul><li>Better art-style and needs less emphasis for rare tags like gynomorph or such other stuff.</li><li>More adaptability with Loras.</li><li>It started to like animating stuff.. Dont ask me why..</li><li>There was an Beta before, What changed?<ul><li>Trained it more as it wasnt as good as I expected</li><li>The Beta made weird images 30% in times</li><li>The Beta didnt even understand what some things even meant, that was weird as I had trained it to do that..</li><li>The new Version has an faster generation (3.8% Faster at some things (at 50 Steps))</li></ul></li></ul>

Files with SHA256: 68536065c9d6e37ec3cdb4f2da350d8aa30d1fe835734cb07d523ca86e74087d