v2.0版更新说明：由于上一版忘记去除训练集图片中的人脸信息，导致生成的画面对人脸会有影响。这一版图片裁切了人脸和头部，生成的画面不会对人脸产生影响。权重最高建议0.9，并且使用full body或cowboy shot等提示词来控制镜头，以免出现人头部脱框的问题。Due to forgetting to remove facial information from the training set images in the previous version, the generated images may have an impact on the faces. This version of the image has cropped the face and head, and the generated image will not affect the face. The highest weight is recommended to be 0.9, and prompt words such as full body or cowboy shot should be used to control the camera to avoid the problem of people's heads falling off the frame.<hr />1、训练集图库素材来源淘宝，lora不包含脸部信息，不会干扰底模或其他lora的脸型，可以随便换脸，如果出现脸部圆斑，请在负面提示词里输入：faceless，并提高权重。1. The training set image library materials are sourced from Taobao. Lora does not contain facial information and will not interfere with the base model or other Lora's facial shapes. You can easily change your face. If there are facial circles, please enter: faceless in the negative prompt word and increase the weight.2、搭配原版Pony模型可能效果不好，建议使用WAI-REAL_CN或WAI-REALMIX等模型，lora强度0.8~1。2. Paired with the original Pony checkpoint, the effect may not be good. It is recommended to use checkpoints such as WAI-REAL-CN or WAI-REALMIX, with a Lora strength of 0.8-1.

Files with SHA256: 01a8896eb6c22057477add719b9e65f4f6d217d29ad8d7a6801978a36ae8b593