The following is translated using ChatGPT.

Training set: screenshots of characters in the game "Honkai: Star Rail"

Trained with NAI.Not sure if overfitting has occurred, will update the training set and try again later.

Weights between 0.7-0.9 will focus more on the character's clothing, while below 0.7 will allow for more variation. Both can be tried.

Additional prompts to use are hair between eyes, red eyes, blue hair

The facial features will be more accurately reproduced.

Some models may produce glasses, but adding glasses as a negative prompt can solve this in most cases.

It may not be suitable for 3D models.

訓練集:崩壞星穹鐵道遊戲中的角色截圖

用NAI訓練，不確定是否有過擬合，之後再更新訓練集試試

權重在0.7~0.9較注重角色衣服，0.7以下比較可以穿脫其他，都可以嘗試看看
關鍵字可以再加上hair between eyes, red eyes, blue hair

臉的部分會比較還原

有些模型可能會出現眼鏡，在負面提示加上glasses可以解決大部分

可能不太適合三次元模型