2.0版本更新:
使用了大约1000张素材,采取大规模的删标训练。
感谢荀帮我处理训练集,辛苦了。
1.9版本更新:
更小的体积,使用DIM4进行训练
1.9版本一共使用了400张精选图片进行训练
触发词:
top_hat,blue_hair,shorts,blue_eyes,thigh_strap,long_sleeves,ahoge,blue_bow,white_shorts,jacket,hat_bow,
Version 2.0 update:
Smaller volume, trained with DIM4.
Version 2.0 used a total of 400 selected pictures for training.
keyword:
top_hat,blue_hair,shorts,blue_eyes,thigh_strap,long_sleeves,ahoge,blue_bow,white_shorts,jacket,hat_bow,
1.5版本:
更小的体积,更好的画面,更自由的动作和换装
此版本克服了1.0版本的动作过拟,拥有更高的自由度,遗憾的是,眼睛的细节似乎没有学到,这将在2.0版本中得到完善。
1.5 version.
The 1.5 version features a smaller size, improved visuals, and more freedom in movements and outfit changes.
This version has overcome the action overfitting issue of the 1.0 version, providing greater freedom. However, it is regrettable that the details of the eyes seem to have not been learned, which will be addressed and improved in the 2.0 version.
训练模型:SDXL1.0官模
测试模型:Kohaku-XL
Training model: SDXL1.0 Official Model
Testing model: Kohaku-XL