sd1.5 v1早期模型，未经过融合或调试，效果较为一般，可以根据需求下载，有概率出现脸崩的情况，因此推荐使用Hires. fix，推荐分辨率7687682，训练和出图底模为anything-v5-PrtRE，sd1.5 v2加入了面部的训练数据，相较于v1面部效果有所改善但仍然推荐使用Hires. fix，加入面部数据外，还根据个人喜好融合了一些别的lora，进行了一定的改善，训练与出图底模为anything v5 prtresdxl版推荐权重在0.8，请勿在0.3-0.6的权重下出图，将会出现很奇怪的图，低于0.3则会出现另外一种很简洁的画风，xl没怎么玩过，这种情况还是第一次见，就很有趣，分辨率大于10241024即可，推荐11521152不推荐过度竖屏或者宽屏的尺寸，容易出现腰部很长的情况，训练底模与出图底模均为animagineXLV31_v31，本模型仅供娱乐，禁止任何商用与违法行为。The early model of sd1.5 v1 has not been fused or debugged, and the effect is relatively average. It can be downloaded according to needs, and there is a probability of face collapse. Therefore, it is recommended to use Hires Fix, recommended resolution is 768 768 2, and the training and mapping base model is anything v5 PrtRE,SD1.5 v2 has added facial training data, which has improved the facial effect compared to v1, but it is still recommended to use Hires Fix, in addition to incorporating facial data, has also integrated some other Loras according to personal preferences, making certain improvements. The training and image base model is anything v5 prtreThe recommended weight for the sdxl version is 0.8. Please do not plot images with weights between 0.3 and 0.6, as strange images may appear. If the weight is lower than 0.3, another very concise style may appear. xl has not played with this before, and this situation is still new and interesting. A resolution greater than 1024 1024 is sufficient. We recommend 1152 1152 and do not recommend excessive vertical or wide screen sizes, as it is prone to long waists. The training and drawing bottom models are both animagineXLV31_v31. This model is for entertainment only and any commercial or illegal activities are prohibited.pixiv：05 - pixiv
