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V1010Dunhuang large model, using a large number of Dunhuang materials for training, and several screening fusion.The following are recommended parameters:vae: 840000Sample: DPM++ 2M SDE KarrasNumber of steps: 20-30Hires: HD fix must be on!!Hd iteration steps :15-20Amplification algorithm: 8x_NMKD-Superscale_150000_G or R-ESRGAN 4x+Redraw range: 0.25-0.4Magnification: 2 or morecfg: 7 (Free to try)

DunhuangMix 敦煌 飞天 古风 

dunhuangmix_v11full.ckpt

v1.1Full

V1.1修剪版

DunhuangMix Dunhuang large model

Connect to model customization and one-on-one teaching QQ Group: 895633778 Telegram: https://t.me/+soU-Ws67aBlkY2I1 For business cooperation, please email: 1269695709@qq.com WeChat: 13353036335 Dunhuang big model, using a large number of Dunhuang materials for training and multiple screening It came from fusion. The following are recommended parameters: va:840000 sampling: DPM++ 2M SDE Karras steps Number: 20-30 Hi-Res Fix (Hires. fix): HD Fix must be enabled!! HD iteration steps: 15-20 Magnification algorithm: 8x_nmkd-superscale_150000_g or R-ESRGaN 4x+ redraw amplitude: 0.25-0.4 Magnification factor: 2 or more cfg: 7 (free to try) Special thanks: Moyu and Majic. The large model of the fusion includes MajicV6, Moyou's full version V1010

Files with SHA256: 5c7e86d6ac47422567094444607edb4823946255bc6ad399bbf48ede027b5506