适配性增强,可以适配男生了，女生可替换面容程度达到75%。使用时权重设为0.6左右。
Enhanced compatibility, can be adapted to boys, girls can replace the face degree reached 75%. Set the weight to about 0.6 when using.
Description
测试版本
FAQ
Comments (8)
效果感覺很棒！等等來試試看。
想請問示例中使用的人物模型？
随便什么大模型 直接跑 效果都还行
效果很棒 可惜對臉型影響太重
有什么建议吗？我现在陷入技术瓶塞
@chocolae 我也不會 我自己訓練的lora也是失敗品
Share the dataset and ideas.
If it has an impact on the face, it may be due to the presence of faces in the training set, reducing the emphasis on facial features in the training set may improve the results.
You mean to delete the character tag and replace it with a girl tag. And then increase the number of pictures without people?
@chocolae Since you have not shared you trained dataset, I can only provide some speculative suggestions. Here are a few ideas to try: 1.Remove all images with people from the training set and retrain. 2.If there are not enough training examples after removing images with people, increase the number of images without people and then retrain. 3.Train separately using images with people and images without people, triggered by different keywords. However, given that your model style does not require people, this idea can be ignored.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.