<h2 id="or-warning-7zu2h1wrt">⚠警告 | warning：</h2><h3 id="nsfw-uwh2t4f2c">本模型禁止用于涩涩，任何相关nsfw内容与我本人无关！</h3><h3 id="attentiondo-not-use-this-model-to-generate-nsfw-pictures.and-all-relevant-uses-have-nothing-to-do-with-me-9c7hkji5r">Attention：Do not use this model to generate nsfw pictures.And all relevant uses have nothing to do with me</h3><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/981d3c3c-c6d0-43ac-8bb7-1da4eafd0511/width=525/981d3c3c-c6d0-43ac-8bb7-1da4eafd0511.jpeg" /><h2 id="or-update-7lgwvj8cn">📜更新 | Update</h2>模型不含垃圾数据无效精度，请放心下载！本模型无需一直开始HI-RES，一直开着相当于浪费时间！本模型专注于描绘可爱娇小的女孩。Model does not contain junk data and invalid accuracy, please feel free to download! It is not necessary to keep HI-RES on all the time. It is a waste of time.KawaICE is a special AI art model that focuses on depicting cute and petite girls.<ul><li>2023.4.26V1与ICE版本更新，这两个版本没有任何的承接关系，V1基于copper模型微调，ICE基于Anything2.1H微调，二者并无关系。因不可抗力，目前V1版本已经下架。</li><li>2023.5.8channel版本更新模型使用Lycoris在底模型上进行dreambooth微调，训练集采用1200张由通配符生成的图片，并使用尸块炼丹法，使得模型即使是由AI生成图训练也能够生成出较为完美的手部动作。<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a42901b1-5d80-4c8e-9ddc-d91f3330f850/width=525/a42901b1-5d80-4c8e-9ddc-d91f3330f850.jpeg" />KawaICE Channel是目前最为好用的幼态体型角色特化模型，欢迎尝试！</li><li>2023.5.9上传了一个作差LoRA，效果实际使用并不好用，还是建议下载ckp模型</li><li>Kawaice的其他版本并不建议使用，效果远不如Channel版本（我自己都在用这个版本）</li></ul><h2 id="or-generation-or0cd85ue">☄️ 生成建议 | Generation</h2>你可以使用您喜欢的任何采样器、步骤、cfgYou can use any Sampler, steps, cfg you like比如我喜欢如下参数：For example, I like the following argument：<ul><li>Sampler: Euler A</li><li>Steps: 20</li><li>CFG: 7</li><li>Clip Skip: 2</li></ul>这个模型不需要非常高的steps就可以出图，采样器选择Euler A/UniPC这种最快的即可，开启HI-RES生成图像并不会有明显的提升，所以无需开启。There are no very high steps required, and turning on HI-RES to generate images will not improve significantly本模型的VAE没有任何问题，无需使用其他VAEThere is no problem to VAE of this model, there is no need to use other VAE<h2 id="or-prompting-3hnztbhfv">🖊️ 提示词 | Prompting</h2>这个模型不需要质量提示词，很多情况下质量提示词会起到反作用This model does not require quality tags.Using quality prompts can backfire,such as<ul><li>masterpiece, best quality,</li></ul>直接描述你想要的效果即可，越详细越好。只使用“1girl”这种提示词抽取图片并不是一个好的方式Simply describe the effect you want, the more detailed the better.Using only prompts like "1girl" is not a good way to extract imagesNegative Prompt Example:<ul><li>(mutated hands and fingers:1.5 ),(mutation, poorly drawn :1.2), (long body :1.3), (mutation, poorly drawn :1.2) , liquid body, text font ui, long neck, uncoordinated body,fused ears,(ugly:1.4),one hand with more than 5 fingers, one hand with less than 5 fingers<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/a058612a-1f96-46ae-99ae-1c293c4df8b9/width=525/a058612a-1f96-46ae-99ae-1c293c4df8b9.jpeg" /></li></ul>负面提示词请不要使用EasyNegative，这会造成模型的部分提示词失效。（下方的图片可能看不出来，但是图1已经明显失去了风格化）Please do not use EasyNegative as this will cause part of the model's prompt words to become invalid.(It may not be obvious from the image below, but Figure 1 has clearly lost its stylization)<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/0841f214-f74f-44f8-a77b-163579881f7f/width=525/0841f214-f74f-44f8-a77b-163579881f7f.jpeg" />【PS：有关EasyNegative以及其他模型的相关内容，可以查看下面的文档】<a target="_blank" rel="ugc" href="https://docs.qq.com/doc/DQ1Vzd3VCTllFaXBv">【模型基础理论】https://docs.qq.com/doc/DQ1Vzd3VCTllFaXBv</a><h2 id="or-specification-4wjcra907">✔说明 | specification：</h2><ul><li>本模型为幼态特化模型，但并非不能出成女等体型角色，如果达不到想要的效果，请加入“small breast”等提示词。</li><li>It is not impossible to produce adult female and other body types. If the desired effect is not achieved, please add "small breast" and other prompt words.<img src="https://imagecache.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/f55f8b00-db22-4e40-2559-d756e2afa400/width=525/f55f8b00-db22-4e40-2559-d756e2afa400.jpeg" /></li><li>本模型在Anything2.1H的基础上进行训练，修改了如“Small chest”、“kawaii”等与模型主题相关提示词的权重。This model was adjusted using the full Anything2.1N model to modify the weights of prompts such as "Small chest", "kawaii", etc. related to the model topic.</li><li>【PS：】请不要添加“masterpiece”等起手式，除非画面出现各种问题Please do not add "masterpiece" or other prompt words</li><li>【四个版本】V1: 效果最差图还是糊的ICE：旧版本Merge：适合融合的版本，如果你想要把这个模型融合到其他模型里面，推荐下载这个版本（没有上传，因为需要的不多，我这里上传太慢）Channel:重新训练的版本，效果最好，推荐下载这个<h3 id="zzzq-w7f2j21wn">之前那个模型因为一些zzzq的原因被干掉了，有需要也不会再发了</h3></li></ul><h2 id="or-other-r1w060kke">❇其他 | other：</h2>为了防止模型被人拿去坑骗小白或者被营销号无脑吹捧，模型禁止转载到任何中国国内模型站或者任何视频平台等，包括单不限于：liblibAI、抖音、小红书、bilibili、炼丹阁等。In order to prevent the model from being cheated or touted by marketing accounts, it is prohibited to upload the model to any model site or any video platform in China, including but not limited to: liblib, 抖音,小红书 ,bilibiliAI, 炼丹阁, etc.模型可以和其他模型一样随意使用，但是请遵守所在地区的法律法规，以免造成麻烦（我们并不负责）The model can be used as freely as other models, but please abide by the local laws and regulations, so as not to cause trouble (we are not responsible for it).<h2 id="-o1c3ljtvq">🥳特别鸣谢：</h2><ul><li><a target="_blank" rel="ugc" href="https://github.com/bmaltais/kohya_ss">GitHub - bmaltais/kohya_ss</a>：训练UI</li><li><a target="_blank" rel="ugc" href="https://github.com/XavierXiao/Dreambooth-Stable-Diffusion">GitHub - XavierXiao/Dreambooth</a></li><li><a target="_blank" rel="ugc" href="https://github.com/AUTOMATIC1111/">Automatic1111</a>: SD-WebUI</li><li>Danbooru：打标器，帮助标准训练集</li><li><a target="_blank" rel="ugc" href="https://github.com/sdbds/stable-diffusion-webui-wildcards">通配符Wildcards</a>：没有这个东西我就不可能整出这样的训练集</li><li>秋叶甜品店：想法！</li><li><a target="_blank" rel="ugc" href="https://civarchive.com/models/50696/qteamix-q">QteaMix：</a>通过研究这个模型得到的思路！</li></ul><h2 id="3-18czx5927">祝大家玩得开心＜3</h2><img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/545249f7-9423-4381-9207-1d96b92e80bd/width=525/545249f7-9423-4381-9207-1d96b92e80bd.jpeg" />

Files with SHA256: f74bd538e4786ccffd0c1effd602b47955c596bc82e2f74f074b027c942853d5