Sign In

Japanese Style Realistic (JSR)

798
8.3k
56
Verified:
SafeTensor
Type
Checkpoint Merge
Stats
7,195
Reviews
Published
May 31, 2023
Base Model
SD 1.5
Hash
AutoV2
51DA760B23
default creator card background decoration
TJ_Max

About JSR V2 and the UPCOMING V3.0:

Because the performance of facial generation in the first version was pretty poor, I merged the popular MajicV5 into V2.0 to make my facial generation more beautiful. However, during subsequent testing, I found that the integration of Majic had too much influence and my model lost its originality (which is probably why I received a one-star rating when it was first released lol..), and the generated faces looked similar and somewhat 2.5D, not realistic enough.

So in the upcoming V3.0, I removed the elements of Majic and made some complex changes to the original merge recipe. In V3.0, you won't see the highly Majic-fied big eyes and the extremely small face, so stay tuned.

因为第一个版本在脸部生成的表现还是比较差,所以在第二个版本里我融合了很火的MajicV5,使得我的脸部生成更加好看及稳定。但是在后续不断的测试中,发现融合majic之后,受其影响太大,我的模型没有了原来的特色(我想这也是为什么在刚发布的时候就收到了一个一星评价T T..),且生成的脸都长得差不多,有点2.5D,不够真实。所以在即将发布的第三个版本里,我把majic的元素去除了,将原来的融合recipe做了一些复杂的改变,在新的版本里你不会看到非常majic化的大眼+瓜子脸,敬请期待

Some sneak peak:

Also, please feel free to share your images in the comments section! Thanks for your support.

也麻烦大家多多在评论区发布自己生成的美图!感谢支持~

About JSR V1.0:

JSR V1 (Japan Style Realistic V1) is trained on over 300 high-quality and high-definition Japanese-style photographic images based on SD, and fused with models such as BRA V4. The original model performs relatively poor in facial generation, but the atmosphere, lighting, and composition of the Japanese-style photographs are excellent. Models like BRA enhance the facial generation and compensate for some of the shortcomings of SD1.5.