All illustrated characters are adults. Please abide by local laws and ensure appropriate content generation.
rurudo Style LoRA
中文 | English 使用前阅读 | Read before use
--

--
仿rurudo的作画风格的LoRA模型。适合画双马尾萝莉。
权重建议值:1 (or 0.6~1.2)
基于模型:Anything-5RE
训练信息:共计130张原画(含裁剪图),最后一次训练迭代520步。
拟合比率:
肢体拟合:12%
人物拟合:98%
构图拟合:30%
技法拟合:102%
其他的一些建议:badhandv4, EasyNegative。
The LoRA model mimics Rurudo's painting style, which is suitable for drawing twin-tailed lolis.
Recommended weight value: 1 (or 0.5~1)
Based on model: Anything-5RE
Training information:
A total of 130 paintings (including cropped images) with 520 iterations.
Fitting ratio:
Motion fitting: 12%
Character fitting: 98%
Composition fitting: 30%
Technique fitting: 102%
Other suggestions: EasyNegative, badhandv4.
--
任何问题,欢迎反馈,问题不提出就不会得到解决。如果你创作出了好作品,欢迎分享它。如果它无法满足你的需求,我是说,我的朋友,请不要给出一星好评。
Any questions are welcome for feedback, as problems will not be solved without being raised. If you have created a great work, please feel free to share it. If it cannot meet your needs, my friend, please do not give a one-star rating.
Description
First version.
FAQ
Comments (7)
大佬,为什么我的质感很差啊?想问一下你用的什么vae
vae-ft-mse-840000-ema-pruned,如果一样的话大概率是炉子不同导致的(图片信息上应该有炉子名)
@Chanter 炉子指的model吗?我用的是yesmix啊
@Chanter 我用你说的那个vae感觉很差。。。。不知道为什么
你看看是不是都能对得上,看着像是超分的问题。Size: 672x1056, Seed: 1169059538, Model: CheckpointYesmix_v16Original, Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 9.5, Clip skip: 2, Model hash: 3e9211917c, Hires steps: 13, Hires upscale: 1.6, Hires upscaler: Latent, Denoising strength: 0.5
@Chanter 谢谢大佬,就是超分的问题,我电脑太差了,一直不怎么用超分。想问一下用超分就会产生这种质变吗?我发现这个vae不用超分根本不行,很多图都必须得超分才行吗?
@2976438468428 第一问:不是用超分就会产生这种质变,用超分的话需要把超分的参数也调好,超分的参数没调好就会出现那种锯齿,可以参考我的参数,超分使用Latent,迭代12~18步,放大倍数1.3~1.8,重绘0.38~0.68,还是不满意的话可以自己琢磨琢磨。第二问:我个人推荐对图片进行超分,不然这个vae的面部容易崩,与其开面部修复,不如直接超分,而且,面部修复通常情况下不咋好使。
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.