


敦煌大模型,使用了大批量的敦煌素材进行训练,并且多次筛选融合而来。
大模型与敦煌lora有什么区别:
1.模型的训练集合更大,更高清,细节还原更好
2.更趋向与真实的质感和光影,dunhuangV3版本为了兼容动漫和真实模型,更偏向于2.8D,衣物流线型更飘逸,各有取舍
3.对真实场景的泛化性更强
4.出图稳定性增强,简单提示加抽卡,可以获得高质量作品,上手门槛更低
5.肢体和脸更稳定
以下为推荐参数:
vae:840000
采样:DPM++ 2M SDE Karras
步数:20-30
高分辨率修复 (Hires. fix):高清修复一定要开!!!
高清迭代步数:15-20
放大算法:8x_NMKD-Superscale_150000_G或者R-ESRGAN 4x+
重绘幅度:0.25-0.4
放大倍数:2或者2以上
cfg:7(可自由尝试)
特别鸣谢:墨幽、Majic,融合的大模型中包含MajicV6,墨幽完整版V1010
Dunhuang large model, using a large number of Dunhuang materials for training, and several screening fusion.
The following are recommended parameters:
vae: 840000
Sample: DPM++ 2M SDE Karras
Number of steps: 20-30
Hires: HD fix must be on!!
Hd iteration steps :15-20
Amplification algorithm: 8x_NMKD-Superscale_150000_G or R-ESRGAN 4x+
Redraw range: 0.25-0.4
Magnification: 2 or more
cfg: 7 (Free to try)
1. The rights to reposted models belong to original creators.
2. Original creators should contact SeaArt.AI staff through official channels to claim their models. We are committed to protecting every creator's rights. Click to Claim
