
Um... Strictly speaking it's not stable (two-person image).
The main training set images come from the anime screenshots of the theatrical version.
Maybe there will be a new version in the future?
I think using the following prompt words as negative may improve the stability of the two-person image.
(blonde hair:0.5),(yellow eyes:0.5),
style
anime screenshot
描述:
训练词语: junta azumaya,takato saijo
名称: dataset.zip
大小 (KB): 122879
类型: Training Data
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success
名称: Takato_Junta_V1-2.safetensors
大小 (KB): 56101
类型: Model
Pickle 扫描结果: Success
Pickle 扫描信息: No Pickle imports
病毒扫描结果: Success