I found a new AI tool Shakker, a best image to image tool. You can try it via https://www.shakker.ai
灵感来源于【瑶圃先春图】和 【会真全图.西厢故事图卷】
人物角色身穿汉服，背景梅花和山水。愿意是想在画面中体现中国古典人物，素雅和飘逸的风格。因为素材图比较少，可能姿势比较单一，大家可以通过controlnet的插件尝试不同的姿势和背景。经过测试，部分达到了想要的效果，权重0.55的情况下可以适配各个主流大模型，配合提示词也可以丰富画面的背景和人物的着装。
大家可以尝试使用 （sd-dynamic-thresholding）这个插件，对于丰富背景和人物细节有很好的效果。也可以尝试clip1 和clip2。
第一次尝试制作模型，有很多疏漏和不足，诚恳希望大家多提意见，帮助我改进！感谢大家！
欢迎大家给我返图！
可以适配各个主流大模型！
推荐大模型：XXMix_9realistic，majicMix系列，
推荐配合使用的lora：墨心、小人书、唐风汉服
推荐参数：0.5~0.7
————————————————————————————————————————
The characters are dressed in Hanfu, with plum blossoms and landscapes in the background. The willingness is to reflect the classical Chinese characters, simple elegance and elegant style in the picture. Because there are few material pictures, the poses may be relatively simple. You can try different poses and backgrounds through the controlnet plug-in. After testing, some of the desired effects have been achieved. With a weight of 0.55, it can be adapted to various mainstream large models, and with prompt words, it can also enrich the background of the picture and the clothing of the characters.
You can try to use the (sd-dynamic-thresholding) plugin, which has a good effect on enriching the background and character details. Also try clip1 and clip2.
This is the first time I try to make a model, there are many omissions and deficiencies, I sincerely hope that everyone can give more comments and help me improve! Thank you everyone!
Welcome everyone to send me back pictures!
It can be adapted to all mainstream large models!
Recommended large models: XXMix_9realistic, majicMix series,
Recommended lora for use with: Moxin, Xiaorenshu, Tang style Hanfu
Recommended parameters: 0.5~0.7
Description
FAQ
Comments (62)
好看~
感谢各位大神的返图，CIVITAI 社区的力量真是太强大了！
Thanks to all the great gods for returning the pictures！ The power of the CIVITAI community is really too powerful !
這個美感太好了，有光劍大大的feel!!
Hi! Guys！ you can try to use this plugin,
https://github.com/Haoming02/sd-webui-vectorscope-cc
There will be unexpected wonderful effects!
大家可以尝试使用这个插件，
https://github.com/Haoming02/sd-webui-vectorscope-cc
在改变色调的同时，会有意想不到的奇妙效果！
感谢大佬，我觉得是古风中最好看的LORA了
这个评价太高了！受不起，受不起！：） 谢谢你！希望可以帮助你生成更多好看的图片！
这个lora还有很多需要改进的地方，谢谢大家多提意见！
美到失语了，玩过的最好看的lora
得闻善言，如逢甘霖！感谢！
中国有句话叫“抛砖引玉”，看看评论区的图片，诚不欺我！！
There is a saying in China that "throwing bricks to attract jade", look at the pictures in the comment area, it is amazing! !
感觉还是你生成的前面的两张比较好看
做出来的效果没有达到各位的 很奇怪 再研究研究
可以根据自己喜欢的风格，搭配不同的lora或者Controlnet的插件，达到自己想要的效果。
谁知道“helloCNPainting”？THX~和这个Lora搭配效果好喜欢~
我也是找不到这个
@648475060ysysys 大佬已经说了，这个模型还在调试~后面会发
是的，aji1 大佬说他在调试中，我也非常期待！
hello,Redpriest,我是Ryan,我很喜欢你的lora，我们有一些商业化的项目想跟你谈谈合作的可能，能否麻烦你加一下我的微信1983219，我们细聊。谢谢
如果是吐司的人，请你滚回去
Wow, this model is truly captivating! The choice of dressing the characters in Hanfu and incorporating plum blossoms and scenic backgrounds creates a beautiful classical Chinese ambiance with a touch of simple elegance and grace. Despite the limited material pictures, the poses and backgrounds are thoughtfully crafted. The experimentation with the controlnet plug-in seems to have yielded some desired effects, allowing for various poses and backgrounds to be explored.
The utilization of the (sd-dynamic-thresholding) plugin is commendable as it enhances the background and character details, resulting in a more enriched visual experience. Additionally, the inclusion of clip1 and clip2 adds a unique flair to the overall presentation.
As a first-time endeavor, this model is already impressive, and any perceived omissions and deficiencies are minor compared to its exceptional qualities. It's heartening to see your openness to feedback and improvement. I'm sure with valuable input from the community, this model will continue to evolve and become even more extraordinary.
Congratulations on this remarkable achievement, and thank you for sharing your work.
Every one can download video from social via https://social-video-downloader.net/
Thank you very much for your approval ！
好模型，值得一试。
说说不好的地方
坏手概率实在是太高
有过拟现象重复姿势很多
原图微博水印没有抠干净
希望作者更优化一下早日更新版本
都是很中肯的建议，确实是有这些问题，我会尽力优化！感谢🙏
繪畫參考用的手部照片素材，不知道對優化有沒有幫助
The Koreans can steal it again and say it's theirs.
The Koreans can steal it again and say it's theirs.
remark Originating from ancient China art,please!
wait what?
麻烦问以下，您的loRA区块权重(Block Weight)中名为PALETTE的预设的数值是多少呀？谢谢
你好！ PALETTE:1,0,0,0,0,0,0,0,0,0,0,0.8,1,1,1,1,1 （角色（去风格化））详细的参数你可以参考麦橘大大的介绍：https://civitai.com/models/43331?modelVersionId=94640
@Redpriest9527 谢谢大佬
@Redpriest9527 感谢，您的模型很棒
@reloadbug274 谢谢！你创作的图片也很漂亮！我很喜欢！
请问大佬有没有交流群，发现很难复刻大佬的图，研究半天不知道差在哪些参数上
well, it looks great but, it's hard to make her smile / sad etc., expression is dim and inflexible
有个疑问，请教一下，参数一致，底模一致，生成出来还是有差异，而且面部崩坏怎么办？
你好，面部如果崩坏，还是要用那个修脸的AD插件
什么时候出更好的版本，坏眼坏手概率很高，张嘴露牙也是
Excellent Lora. The ancient Chinese style, color and mysterious atmosphere is fascinating. Just need a bit improvement of fingers. This Lora might increase mutated hands and fingers problems. Still one of the best Loras.
Thank you very much for liking this Lora. Your suggestions are very helpful to me. grateful!
这个作者很有意思，提供了一个LORA，下方的出图区除了他自己没有一个能复现他这种效果的。可以说是中国传统的教徒思维，永远留一手
SD1.5时代，单靠一个LORA直出各种姿态确实面临许多挑战，尤其是表现中国古风的各种飘逸美感。因此，社区发展出了多种方法来优化效果。例如结合OpenPose，或多个LORA组合使用，甚至分层调用参数等，方法非常多样化。我也希望大家各显神通！能够找到适合自己的使用方式！
关于您提到无法复现我的效果，可能是因为最终图像不仅依赖于LORA本身，还受到CN插件中许多功能的影响，而这些插件 的使用参数最终是不会体现在图片的信息数据里的。如果您希望获得相似效果，也许需要调整更全面的参数组合。
感谢您的反馈！(今天是2025年5月4日，社区已有众多优秀模型可供选择，您也可以尝试其他更适合您需求的资源)"
你这是无能狂怒还是羡慕嫉妒恨
大佬您好，我想请教一下如何训练衣服纹样的lora。您的lora衣服质感和纹样也太好看了
Details
Files
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.