SilvermoonMix03-Chenkin-Rectified-Flow focuses on natural lighting, performing exceptionally well in low-light environments and high-contrast scenarios. It is specifically tuned for abstract elements and fluid textures, such as hair, water, ribbons, etc.
For instructions on how to use the RF model, please refer to https://civarchive.com/models/2363696/chenkinnoob-xl-v02-rectified-flow
I mainly use ReForge. Support for RF in ReForge is being implemented through a built-in extension:
Why RF?
ChenkinNoob (both the original and RF versions) features an updated new knowledge base (up to 01/2026). However, the original ChenkinNoob was based on E-pred, which doesn't perform well on dark contents and has an issue where it tends to fill the entire picture—a stylistic trait I wanted to avoid.
The Rectified-Flow (RF) version resolves both of these issues.
The fine-tune result on the ChenkinNoob RF demonstrates good scene and limb logic. The raw generated results are already acceptable without using Hires. fix or ADetailer,
Artist tags perform closer to the artist's true style compared to the Noob-Vpred version. You can expect an approximate 20%-25% improvement in style accuracy.
Recommended settings:
Sampler: Euler A Comfy RF (for ReForge)/Euler
Steps: 20-28
CFG: around 5
Shift: around 3
Schedule: Normal / Simple / SGM Uniform / Beta
Positive Quality Tags:
masterpiece, best quality, aestheticNegative Tags:
worst quality, normal quality, bad anatomy, low resolution
Note: All my demo images were generated via stable-diffusion-webui-reForge. You can right-click any of the demo images -> open in a new tab to download the original image. You can then pull it into the webui's PNG info tab to see my exact generation settings.
SilvermoonMix03-Chenkin-v1-RF 专注于自然光影，在低光环境和高对比度场景下表现出色。它针对抽象元素和流体纹理（如头发、水、丝带等）进行了专门优化。
有关如何使用 RF 模型的说明，请参考： https://civarchive.com/models/2363696/chenkinnoob-xl-v02-rectified-flow
我主要使用 ReForge。ReForge 对 RF 的支持正通过一个内置扩展来实现：
为什么选择 RF 版本？
ChenkinNoob（原版和 RF 版）都拥有最新的知识库（截至 2026 年 1 月）。但是，原版 ChenkinNoob 是基于 E-pred 的，它在处理暗部内容时表现不佳，并且倾向于把整个画面填满，我不喜欢这种风格。
RF 版本完美解决了这两个问题。
此外，在ChenkinNoob RF 上的微调结果展现出了非常优秀的场景和肢体逻辑。即使不使用 Hires. fix（高清修复）和 ADetailer，直接生成的原图效果也已经非常不错且完全可以接受了。
画师 tag 的表现更加贴近原画师的真实风格。与 Noob-Vpred 版本相比，画风的还原度大约提升了 20%-25%。
推荐设置：
采样方法 (Sampler): Euler A Comfy RF (ReForge自带)/Euler
迭代步数 (Steps): 20-28
提示词引导系数 (CFG): 5 左右
Shift: 3 左右
调度器 (Schedule): Normal / Simple / SGM Uniform / Beta
正面质量提示词:
masterpiece, best quality, aesthetic负面提示词:
worst quality, normal quality, bad anatomy, low resolution
注意： 我所有的示例图片都是使用 stable-diffusion-webui-reForge 生成的。您可以右键单击任何示例图片 -> 在新标签页中打开以保存原图。然后，您可以将其拖入 WebUI 的 PNG Info 选项卡中来查看我的具体生成参数。
Change notes:
v2.1: Fine-tuned on ChenkinNoob-XL-v0.3 Rectified-Flow.
v1.0: First release. Fine-tuned on ChenkinNoob-XL-v0.2 Rectified-Flow.
Description
v1.0: First release. Fine-tuned on ChenkinNoob-XL-v0.2 Rectified-Flow.
FAQ
Comments (45)
Hi Boss, Have you by any chance tried this on Forge Neo or Classic?
I'm struggling with Comfi—I can't figure it out; it's too complicated for me XD
Hi @mifink94 I use ReForge. It is supported out of the box.
No matter which version of Forge you use, NAI runs perfectly fine. Are you experiencing this issue?
@BIG_A Nono, theres no problem yet, Im just confused by this sampler. And Neo doesnt have that extension. Need to try it
@mifink94 I ve tried this checkpoint on Forge classic AND neo, It DOES'T WORK~~you are right~
@BIG_A thenks bro)
We have three options:
Download reforge
Download comfi
Ask the Neo developer to add it (but most likely, he simply deleted it from there)
@mifink94 I agree the last one!
@BIG_A I wrote to him anyway, just in case)
@mifink94 well done!
Hello, I have seen the new model, thank you for the many models you always provide. The description says that it performs better than your SilvermoonILEvolved in some aspects.
I like the sample images you've created, and it also seems to have a fairly up-to-date knowledge base. Can it already create Endfield characters well?
My main question is: How does it handle complex prompts? Your SilvermoonILEvolved is well suited for complex prompts and, in my opinion, is one of the best SDXL-based models, probably because it is vPred. Now I'm wondering how good this one is in complex scenes, with multiple characters without inpainting or regional prompting, and long complex poses?
And how does it perform when using Lora?
Would you say that it is generally better than SilvermoonILEvolved in all aspects, or does it simply have its own strengths?
I can only generate online on Civitai, so I have to bid on it and then I can use it on Tuesday or later.
Thanks for your great contributions/models as always.
I wanted to ask the same thing, but you beat me to it)
@mifink94 @VKilko I haven't had a chance to compare with other models side by side. I simply feel the "good image rate" is even higher than my SilvermoonMix Evolved when generating demo images.
I'm not sure whether the civitai online generator supports RF models. RF models require a sd3 node in comfy and a advanced sampling extension in ReForge.
No one has tested this yet. I don't want to waste your yellow buzz on bidding.
@silvermoong I asked the support and wait for an answer.
You say the good image rate is better then evolved. And you don't compared it.
But the question how good the model with complex prompts and multiple characters is, do you ca test it please? But don't feel any pressure or so, you don't need, I only want to now for the case that RF works onsite?
And what is with your evolved series? Do you make an evolved RF model? And how good are the RF with Endfield-characters and Lora? Can anyone answer this now?
@silvermoong all is good, boss, I just weiting when in Fneo add this sampler)
Im really happy to see you upload another great model ^_^ This works very well!!
0.3 RF should be out soon and I wonder how it would match with the evolved version. This is awesome though thanks for trying RF.
Also upping the sd3 shift can help at least with 0.2 rf. I saw Anzhc suggesting 4.5
@HaloSkull Thank you for the suggestion. I haven't had much time playing with shift. Through my quick test on SilvermoonMix 03 RF, lower shift gives better composition, higher shift brings better details. I choose shift 2 after a few tests.
Since my other comment got deleted. 0.3 RF should be close to done. Also RF 0.2 & 0.3 should be better than regular chenkin because Blu & Anzhc fixed the tags that were messed up.
想问一下大佬下了个reforge版本的webUI，为什么出的图还是色块啊
ReForge内置插件Advanced model sampling，要选sd3
hello boss, it looks very impressive just one question. what is the value you use on shift (SD3)? The default value is 4.5 but the chenkinNoob guide says 2.5. I find 4.5 to give better results. Thanks!
@Rehvka It is really up to you. I use 2
@Rehvka Hey, I played with the shift a bit more. 3 might be a sweet point, and >5 might be a bad idea.
@silvermoong ty for the update! Im not really sure what Im doing wrong but all of my test have been BAD ;-;, either too washed or too fried. Ill try your set up later, thanks!
Looks like I'll have to download Reforge to follow in the master's footsteps; apparently, Forge Neo doesn't support it.
@NomadSterben Ideally, I’d build a model that just works out of the box. But I’ve found that extra setting of using Rectified Flow is a necessary trade-off to meet my quality expectations.
@silvermoong So I finally got the chance to try your RF work. Reforge had a bug with CLIP installation, but Forge Neo added support a few days ago.
The aesthetics and overall image coherence are really good, but the hands… dear god, send help. ADetailer also misses sometimes like a broken shotgun 🤪
Euler A Comfy RF - where do get this sampler ?
you get it from the WebUI Forge/reForge, a stable diffusion fork
Excellent! Now we await the RF version of Evolved...
@wtre59 Chenkin RF v0.3 might be released soon. I will see how that performs.
@silvermoong March 10th for 0.3 RF
If evolved gets a RF version, doe it also gets a VPred version? RF don't works fully on the site. "Unfortunately, if there are special nodes required in order to generate with the checkpoint, we are unable to support on-site generation for the model at this time."
Hi @VKilko, If I train an RF version of SilvermoonMix Evolved, it will utilize a training strategy similar to the v6.0 Vpred release. Consequently, there are no plans to upgrade the existing SilvermoonMix Evolved further at this time.
用adetailer脸部会直接黑掉 这是为什么
@lingmo 你有用reforge专门的ADetailer吗？https://github.com/Anzhc/aadetailer-reforge
For me, Adetailer usually ruins everything🤣
hi boss) you generate in 832*1216 and then upscale by 50% right?
@mifink94 Yes
@silvermoong share settings of hrfix please🙏
@mifink94 You can right click any of my demo images, open in a new tab and save the image. This gives you the original image with generation metadata. You can load my generation settings using the webui's PNG info tab.
@silvermoong oh, really, thank you a lot)
hi there, thanks for this beautiful model!
Does IL/NOOB Esp LoRAs have any effect, if used with this model?