短暂复活其实和1.0没什么关系,只不过懒得分一个新页面了,我不知道有多少人跟我一样喜欢传统艺术和印象派,示例图基本上都是靠艺术类tag做的,其实1.5对这些传统艺术家的支持是挺好的都不需要lora就能实现和动漫的完美融合,其实在ai刚开始有苗头的时候我就已经在做这类尝试了,当时就觉得十分的惊艳
触发词:post-Impressionist (只是我个人比较喜欢用的艺术类tag)
其实可以输入任何你懂的艺术流派或者知名艺术家,不过这得看sd1.5底模有没有训练过这个艺术家的画,我实际上做的工作就是把模型调整得好看点然后通过原版训练集里的艺术家支持让画面看起来很好,当然你不加艺术家tag原版的模型也是挺好用的
艺术家名单
SD1.5 Artist Study - Google 表格
The brief resurrection has nothing to do with 1.0, but I'm too lazy to create a new page I don't know how many people like traditional art and Impressionism like me. The example images are mostly made by art tags. In fact, 1.5's support for these traditional artists is quite good, and they don't need Lora to be perfectly integrated with anime
trigger:post-Impressionist (It's just an art tag that I personally prefer to use)
Actually, you can input any art genre or well-known artist you understand, but it depends on whether the SD1.5 base mold has trained this artist's painting. What I actually do is adjust the model to look better and use the artist support in the original training set to make the image look good. Of course, the original model is also very useful without adding an artist tag
SD1.5 Artist Study - Google 表格
尝试做了一下XL的模型太卡了,而且非常的不顺利所以放弃了
I tried to make the XL model. It was too difficult, so I gave up
Description
FAQ
Comments (23)
方便说一下封面的tag吗,被封面吸引进来的
1girl,birch forest,monet color,(antler:1.2),naked,stream,close shot,huge chest,close shot,anime characters,white hair,deer hair,the whole body,oil painting,post-impressionist,(face upward:1.2),Post Impressionist,Droplets,well,soft lighting,
负面:easynegative,bad-hands-5,(worst quality, low quality:1.4),lowres,skin blemishes,extra fingers,fewer fingers,strange fingers,Hand grip,(lean),Strange eyes,(three arms),(Many arms),(watermarking),badhandv4 deformityv,Monster's Fingers,ankylodactylia,Twisted Fingers,Fingers with excessive joints,Extra Long Fingers,Multiple nipples,Excessive areola,skirt,meteor,strange lines,KHFB AuroraNegative
awesome man
Hello,
Is there a VAE that needs to be used to replicate above images?
Also what happened to LongSeason model?
Maybe it would be better to add a highly saturated VAE
大佬的模型风格我都挺喜欢的
从一个图片的模型链接,偶然找到这个大模型,初看以为只是一个普通插画模型,多看几张图片,发现如此经验,光影,色彩,令人赞叹不已,不仅仅是作者工作程度,更佩服其审美水平。颇得印象派精髓,作为曾经的绘画工作者,深深像你致谢,让我虽然放下画笔,任然可以重拾印象派的留恋的审美。期待你的sdxl 新模型
非常喜欢印象派在景物上的出色表现!不过还有一点问题,我用与例图完全一致的prompt和配置输出的图像却色彩发灰,与例图的饱和度完全是两样,作者有头绪吗?
加载84000vae
大佬你太牛逼了!!!我下载了上百个模型,仔细分类测试,你这个模型前无古人,真是太漂亮了,感谢大佬的无私分享!!!
大佬的所有模型我都准备下载了
搞这些的人太缺少审美和艺术方面的认知了,我是搞美术的所以稍微有点这方面的知识,很多人就乱融,然后融出来也不怎么搞测试就发,导致10个模型8个过拟合,要我说真没必要下那么多模型留几个顶部模型和几个风格独特但是不过拟合的就行了,我刚开始也特别爱猛猛下模型,直到有一天我测试模型的时候3个顶流模型出的图都一个样我就自己搞了,垃圾数据是真多没必要浪费流量空间下那么多模型
@Livesting 说的非常对,相互融毫无意义
二维,2.5模型中颜色最牛的模型,我二维图片最后都希望用他来放大,可以呈现出同色调中最漂亮色彩的图像,我的大爱。
感激涕零
是天使...!!!
非常非常好的模型,色彩非常丰富。
非常喜欢这个模型,光影和颜色都非常好。希望有后续作品。
Can you make the pony model? it will be awesome
What hires fix settings should we use and how do you upscale with this model? When I upscale or do hires fix it comes out with a lot of noise over the image.
I think it depends on the GPU? but usually for SD1.5 based models the denoise is around 0.50 minimum. Any lower and you get artifacts.
There are ways you can constrain the image though if the upscale is changing stuff too much. You can do control net depth, tiles etc, all of them will help maintain the general input
Its beautyful checkpoint for mariners and very cool portraits. Thank you, author! <3
Absolutely beautiful.
Details
Files
exquisiteDetails_art.safetensors
Mirrors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
Exquisite.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
exquisiteDetails_art.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.





