Translated description:The model idea comes from LOL Titan Peak_Tianjie old version background illustration,
Recommended base models: First release on the entire network|SHMILY Classical Colorful_v1.0,The model address is here:https://huggingface.co/swl-models/shmily-GuDianXuanCai-v1.0/tree/main,pasteldiffusedmix_v22,
VAE recommends opening: vae-ft-mse-840000-ema-pruned.safetensors
Sampler recommendation: DPM++ 2M Karras,
Recommended number of iteration steps: 30
It is recommended to turn on high-definition restoration: the amplification algorithm is R-ESRGAN 4x+ Anime6B, the amplification is 2 times, and the redraw amplitude is 0.3
ADetailer is turned on, AD model recommendation: face_ yolov8n.pt
Lora weight recommendation is 0.5-0.75
CFG is: 5.5
CLIP recommendation: 2
The generalization is very good and can be combined with other loras. I hope everyone will return the pictures and like them more. Thank you.
模型创意来于LOL巨神峰_天界老版背景插画,
底模推荐:全网首发|SHMILY古典炫彩_v1.0,模型地址在这:https://huggingface.co/swl-models/shmily-GuDianXuanCai-v1.0/tree/main,pasteldiffusedmix_v22,
VAE推荐开:vae-ft-mse-840000-ema-pruned.safetensors
采样器推荐:DPM++ 2M Karras,
迭代步数推荐:30
推荐开启高清修复:放大算法为R-ESRGAN 4x+ Anime6B,放大2倍,重绘幅度为0.3
ADetailer开启,AD模型推荐:face_yolov8n.pt
lora权重推荐0.5-0.75
CFG为:5.5
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