Muggle's Hotpot_SDXL

"If you have custom modeling or other cooperation and consultation needs, please contact Wechat: pomelonioi. Please briefly state your purpose. I appreciate your understanding as time is valuable and prefer not to engage in casual chats. Thank you!

Hello everyone, I'm a Muggle! After more than 30 failed attempts at different versions, I've finally developed an SDXL model that has excited me for a long time.

Based on HelloWorld_v2.0 (special thanks to LEOSAM ), Hotpot_SDXL has undergone intensive training in two areas: 'Lighting' and 'Physique.'

In terms of lighting, the material enhancement has played to the strengths of SDXL, bringing professional-grade image quality. The shadows on the face, backlighting on the hair, and contours of the person all have a high-end photographic feel.

For physique, the material supplements the weaknesses of SDXL, enriching the postures and movements of characters. Full figures, long legs, and model-standard poses significantly enhance the full-body image capability.

These two types of materials are like a yin-yang hotpot, one side being elegant and refined like a spring snow, and the other being down-to-earth and folksy, most suitable for depicting the beauty of Chinese women with black hair and eyes, hence the name Hotpot — Muggle Hotpot.

Here are some tips for using the model:

1. The trigger word 'mugglelight' can be used as a quality term in your initial approach.

2. My personal page has uploaded two new sets of wildcards: 'Muggle Fantasy Lighting' and 'Muggle Spring Night.' These are collections of prompt words I consider good, and they enhance the effect when used with the model.

3. SDXL has better natural language recognition, so when writing prompts, I no longer recommend the single-word method completely used by sd1.5. You can refer to my example images or decks, and use 'a sentence to describe the main theme of the picture + multiple words to supplement details' for a better effect.

4. This is a specialized model for Asian women, and its representation of European, African, and male subjects needs to be improved.

5. Combine with turbo & LCM lora for rapid image generation.

6. Example parameters: sampler restart, resolution 832*1344 (1.5x high-definition restoration, adjust down if it causes memory overload), CFGscale 7-9 (or use the Dynamic Thresholding -CFG Scale Fix plugin). For full-body or long-distance shots, consider turning on adetailer, and in the plugin, you might fill in: detailed face, detailed skin. (These aren't mandatory settings, just my personal preference.)

7. The VAE model is already built-in, choose none or automatic, and do not use sd1.5's VAE such as vae-ft-mse-840000 as it will cause image distortion!

If you find the model useful, please don't forget to like and share your images back with the model, it's really important to me!

Additionally, for topics on model training, raw image prompts, or using the card-drawing machine (automatic image generation), feel free to join my discussion group at 891417811."

Model Usage Statement

1. You may not host this model or its derivative versions (such as integrated model versions) on websites/applications intended for income generation or donations. If you wish to use this model or its derivatives for commercial purposes (such as generative services, selling images, using images for articles or publications on short video platforms, etc.), please contact me through the platform.

2. You are not permitted to directly sell images generated by this model unless you have made substantial manual modifications to the image, to the extent that it can legally be considered your own personal work. Should you violate this term, all legal consequences are your sole responsibility, and I hereby disclaim all liability.

3. You must not use this model to intentionally create or share illegal or harmful content. Please adhere to public order and good morals and utilize this model for positive and constructive purposes only.

如您有模型定制或其他合作、咨询需求，请联系Wechat: pomelonioi。请简单注明来意，时间宝贵不闲聊哈，谢谢理解！

大家好，我是麻瓜！经过30多个版本的失败尝试，我得到了一款让我兴奋许久的SDXL模型。

基于HelloWorld_v2.0（特别鸣谢：LEOSAM），Hotpot_SDXL在【光影】和【肢体】两方面做了加强特训。

光影类素材增强了SDXL的长处，带来了很专业的画面质感，脸上的阴影、头发的背光、人物的轮廓都有着很高级的摄影镜头感；

肢体类素材则补充了SDXL短板，丰富了人物的姿势和动作，丰满的身材、修长的双腿以及模特标准的姿势，大大增强了全身图的表现能力；

这两类素材恰似一口鸳鸯火锅，一边是高雅唯美的阳春白雪、一边是接地气的下里巴人，且最适合描绘黑发黑眸的中国美女，故而得名Hotpot——麻瓜火锅。

以下是模型的使用贴士：

（1）触发词mugglelight，可以当作质量词纳入您的起手式；

（2）个人主页上传了【Muggle梦幻光影】和【Muggle春宵时刻】两套全新的wildcards，即我认为不错的提示词合集，配套模型使用效果更佳；

（3）sdxl对自然语言识别更好，写prompt的时候不再推荐sd1.5完全使用单个单词的写法，大家可以参考我的例图或者卡组，用“一句话描述画面主题+多个单词补充细节”的写法，效果更佳；

（4）亚洲女性特化模型，对于欧美非洲以及男性的表现有待增强；

（5）配合turbo&LCM lora可实现急速出图；

（6）例图参数：采样器restart、分辨率832*1344（高清修复1.5倍，爆显存可以不开或者调小）、CFGscale 7-9（或者使用 Dynamic Thresholding -CFG Scale Fix插件）

全身图或者远距离镜头建议开启adetailer，插件中可填写：detailed face,detailed skin

（并非一定要这样设置，只是我的小小个人习惯）

（6）VAE模型已内置，选择none或者自动就好，千万不要使用sd1.5的vae比如vae-ft-mse-840000会导致花图噢！

如果您觉得模型好用，请不要忘记给模型点赞+返图，这对我真的很重要！

另外，关于模型训练、生图prompt、抽卡机（自动生图）使用等话题，欢迎加入我的交流群891417811

模型用途声明

1. 你不得将此模型及其衍生版本（如融合模型版本）托管于计划赚取收入或捐赠的网站/应用程序。如果您需要将此模型及其衍生版本用于商业目的（如生成式服务、售卖图片、将图片用于公开发表到短视频平台的文章或出版物等等），请通过站内联系本人

2. 您不得直接售卖由此模型生成的图片，除非您对该图片进行了足够程度的人工修改，使其在法律意义上可以被完全判定为您的个人作品。如果您违反本条，所造成的一切法律后果由您个人承担，请恕本人概不负责。

3. 您不能使用该模型故意制作或共享非法或有害的内容传播和输出，请您遵守公序良德，将此模型用于积极正面的用途。