Type | |
Stats | 2,006 85 |
Reviews | (235) |
Published | Aug 1, 2023 |
Base Model | |
Training | Steps: 9,820 Epochs: 10 |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 1ED28AA3F8 |
The V2.0 model used more than 400 high-quality original pictures for training.
This LORA model was originally prepared for SDXL1.0, so the original training materials are all above 1024 resolution.This V2.0 model has very good details, including the material texture on the pants, which can be restored very well (please enlarge the picture to enjoy)
The LORA model of SDXL1.0 version is still being iterated, and there is no particularly satisfactory effect yet, and it will continue to be updated in the future.
Unexpected blockbuster effects can be obtained with HI-RES.
Recommended output resolution: 1024 x 1344
Recommendation Weight：0.6~1.0,
Recommendation VAE：animevae.pt
Recommendation Model：scmix_NSM.ckpt
--- --- --- --- --- --- --- ---
该V2.0模型使用了超过400张高质量的原始图片进行训练。
这个LORA模型原本是为 SDXL1.0准备的，所以原始训练素材均为1024以上分辨率。V2.0版本拥有非常好的细节表现，包括裤子上的材质纹理都能够还原的非常棒（请放大图片欣赏）
SDXL1.0版本的LORA模型还在迭代中，目前还没有特别满意的效果，后续会继续更新。
使用高清修复可以获得意想不到的大片效果。
建议出图分辨率：1024 x 1344
推荐权重：0.6~1.0
推荐vae：animevae.pt
推荐模型：scmix_NSM.ckpt