The main function of this lora is to assist models that cannot draw a male image to output a male image,and make men’s faces cleaner,Everything else is incidental.If your face is blurry, please reduce the redrawing range of the Face Editor plug-in.If you are pursuing better picture quality, please try to refer to the negative tags I provided.It is recommended that the full body picture be 720x1280, and then enlarge it in high definition.这个lora的主要功能是辅助绘制不出男性形象的模型输出男性形象，并且使男人的脸更干净，其它的都是附带的。脸糊请把修脸插件的重绘幅度调小。如果追求更好的画面品质请尽量参考我提供的负面tag。全身图建议720x1280，再高清放大。详情往下翻↓↓↓。Effect:作用：1. Produce a positive tendency of male aesthetics, improve male appearance, increase alpha temperament, based on Asian female aesthetics.1.产生男性美学的正向倾向，提高男性颜值，增加阿尔法气质，基于亚洲女性审美。2. Reduce non-2D models by 0.5 dimensions.2.把非2D类的模型都减少0.5个次元，3. Increase the probability that male images can appear stably in all models,3.提高在所有模型中男性形象能稳定出现的概率，4. Provide data sources for some formal suits,4.提供部分西装正装的数据源，5. Strengthen the stability of men's whole body and half body (mainly standing posture).5.加强男性全身半身(主要是站姿)的稳定性。6. Add scene details and improve the beauty of the composition.6.增加场景细节，提高构图美感。Recommended prompt：1male,mature male,solo,male focus,slender eyes,low brow,square jaw,formal,collared shirt,necktie,open jacket,suit,pants,broad shoulders,You can delete it according to your needs.Feature：hair slicked back,curtained hair,gold-rimmed glasses,side-split bangs,swept bangs,Composition：full body，cowboy shot,upper body，Shoulder jacket：wear overcoat over suit,coat draped over shoulders without putting arms in the sleeves,jacket draped over shoulders without putting arms in the sleeves,empty long sleeves,(Usually in order to wear a normal suit stably, you need to put the "shoulder jacket", which is the ↑ tag above, on the negative side)Clothing details：open jacket,sleeves rolled up,open collar,shirt tucked in,A few months ago, my female college classmate wanted me to use AI to draw "Domineering CEO". However, if you play with AI painting, you will know that most of the base models are seriously effeminate, making it difficult to generate a customized adult male image. . So when I decided to learn how to train models, I made training "Domineering CEO" my first project. I seriously underestimated the difficulty of getting all the base models to draw men, and it took 4 months.The Sun version is the first excellent model that emerged after 2 and a half months of groping in the maze of models. Its art quality and physical stability have never been surpassed by subsequent models. Its characteristic is that no matter what kind of man is drawn, it is Bright justice. It is more like a gender-reversed version of the female figure in the base mold, with less impurities than the later version.So in order to meet the characteristics of "Domineering CEO", I spent more than a month making the Alpha version and the Moon version respectively.The data volume of the Alpha version and the Moon version is more than twice that of the Sun version. After more than 4 major adjustments, the tag accuracy will be higher and the posture will be richer than the Sun version. The disadvantage is that as the amount of data increases, the difficulty of grasping the style direction increases exponentially. So in my opinion, the art quality performance is not as stable as the Sun version.The alpha version of the man has thicker hair and a burlier figure. The Moon version of the man has squinty eyes, the atmosphere of the scene is gloomy and dull, and the overall performance is a bit over-fitting, It is more suitable for 2D style models, and the recommended weight is lower than that of Sun and Alpha. The training sets of Alpha and Moon are 90% identical. The data of the Moon version is fresher, but the performance is not optimal. So just choose to download according to your preference.Precautions:Faces may crash on some models, requiring a plugin.It is recommended that the full body picture be 720x1280, and then enlarge it in high definition.Since the training set images are in 2D style, models containing real models above 2.5d will be reduced by 0.5 dimensions.Half of the training set of this model has been corrected by my own manual painting. It is only used for learning and communication, and commercial use is prohibited.This model, improved/fusion models based on this model, and its products are strictly prohibited from any commercial use.This is my first lora. Excellent training sets are difficult to obtain. It took nearly 4 months and I did everything I could.Welcome to leave a message to feedback questions and make suggestions, and we will take time to make up for the shortcomings.My English is very poor and I used translation software. I apologize for the deviation in semantic understanding.推荐prompt：1male,mature male,solo,male focus,slender eyes,low brow,square jaw,formal,collared shirt,necktie,open jacket,suit,pants,broad shoulders,可以根据自己需要删减。特征推荐：hair slicked back,curtained hair,gold-rimmed glasses,side-split bangs,swept bangs,构图视角：full body，cowboy shot,upper body，肩上外套：wear overcoat over suit,coat draped over shoulders without putting arms in the sleeves,jacket draped over shoulders without putting arms in the sleeves,empty long sleeves,(通常为了稳定的跑出正常的西装正装，需要把“肩上外套”也就是↑上面这段tag写在负面)服装细节：open jacket,sleeves rolled up,open collar,shirt tucked in,数个月前，我的大学女同学希望我跑霸道总裁，但如果玩AI绘画的话就会知道大部分的底模都被严重娘化，很少有能跑出男人的，更何况是这种定制向气质的成年男性形象。所以当我决定炼丹的时候，我把炼霸总丹作为了我的第一个课题，我严重低估了让所有底模跑男人这一需求的难度，花了4个月时间。Sun版本是在丹之迷宫中摸索2个半月后出现的第一枚神丹，美术质量和身体稳定性，后面再炼的丹再也没有超越过，特点是不管跑什么样的男人都自带圣光，一身正气。它更像是底模女性形象的性转版，比后面的版本杂质更少。所以为了符合霸总特质，我又花一个多月时间分别调制阿尔法版本和Moon版本。Alpha版本和Moon版本的数据量是Sun的2倍以上，经历了4次以上的大调优，tag正确率也会更高，姿态比Sun更丰富。缺点是由于数据量变大，风格方向把握难度成倍上升。所以美术品质表现不如Sun稳定。Alpha版本的男人发量多一点，身材更魁梧一些。Moon版本的男人有吊梢眼，场景氛围阴郁单一些，整体表现有点过拟合了，更适合2D向底模，推荐权重低于Sun和Alpha。Alpha和Moon的训练集雷同度达90%，Moon版本的数据更新鲜些，但表现不是最优。所以看自己喜好选择下载即可。注意事项：脸部在部分模型上可能会崩溃，需要插件。由于训练集图片为2D风格，所以2.5d以上含真实系模型都会被减少0.5个次元。此模型半数训练集经过我本人手工绘画纠正，仅作为学习和交流使用，禁止商业用途。此模型及基于此模型的改良/融合模型，以及其产物严禁任何以盈利目的的使用。这是我的第一个lora，优秀的训练集很难获得，耗时近4个月，我做到了我能做到的一切。欢迎留言反馈问题，提出建议，会抽时间来补足不足之处。

Domineering CEO

Files with SHA256: c3397f880161aa616df2738464f069e40a5595a09d2745ee2d6719cc5efe41b8