This model training set is designed by Overwatch people to train, can produce science fiction, mecha, fashion style of European and American game character design.
Trigger word: HEZI
Default prefix: HEZI,game character design,Overwatch style,European and American character design,
Bottom mold recommendation: The whole net first | SHMILY classical colorful _v1.0,
Sampler recommendation: DPM++ 3M SDE Exponential
Recommended number of iteration steps: 30
It is recommended to open HD repair: the amplification algorithm is 4x-UltraSharp,
Open the ADetailer, recommended AD model: face_yolov8n.pt, recommended to open an ad model for repairing eyes: mediapipe_face_mesh_eyes_only
The recommended weight of lora is 0.7
CFG: 5 to 7
CLIP Recommended: 2
The generalization is very good and can be combined with other lora, I hope you can return more pictures and praise more, thank you
This model is only used for communication and learning, and is not commercially available. It is open to download on various platforms. If it is integrated, please mark the original model to avoid unnecessary trouble
此模型训练集是由守望先锋人设进行训练，可以出科幻，机甲，潮流风格的欧美游戏人物设计。
触发词：HEZI
默认前缀：HEZI,game character design,Overwatch style,European and American character design,
底模推荐：全网首发｜SHMILY古典炫彩_v1.0，
采样器推荐：DPM++ 3M SDE Exponential
迭代步数推荐：30
推荐开启高清修复：放大算法为4x-UltraSharp，
ADetailer开启，AD模型推荐：face_yolov8n.pt,推荐多开一个修复眼睛的ad模型：mediapipe_face_mesh_eyes_only
lora权重推荐0.7
CFG为：5~7
CLIP推荐：2
泛化性很好可结合其他lora，希望大家多返图多点赞，谢谢
此模型仅用作交流和学习，不可商用，在各个平台开放下载，如果融合请标注原模型以免不必要的麻烦