NOTICE:LET ME KNOW before you put this model on commercial usage.
My twitter account:@eagelaxis :) Contact me if needed.
Discord Account:Eagelaxis#7818
Version Choosing Advice:V2f,V3,Coda and V3.5 are recommended,especially CODA for first-time users.
Hard to tell how many models used to merge.
Check the example images to recognize this model's art style
For more example images, just take a look at https://pixai.art
More attention on shades and backgrounds compared with former models(Andromeda-Mix | Stable Diffusion Checkpoint | Civitai)
Hands-fix is still waiting to be improved.
Highres-fix(upscaler) is strongly recommended(using the SwinIR_4x,R-ESRGAN 4x+anime6B by myself) in order to not make blurry images.
(Sorry for the misunderstanding caused by myself."Bad_prompt_v2:1.4" is a kind of embedding recognized as collection of normal negative prompts such as"lowres,bad anatomy".)
Recommend: Clip skip 2 Sampler:DPM++2M Karras Steps:20+
CFG scale:4-8 Vae:Pastel-Waifu-Diffusion.vae.pt(The vae used by Pastel-mix si just good enough)
Highres.fix:SwinIR_4x Hires steps:10+ Denoising strength:0.4+ Upscale by: 1.5+
Loras along with embeddings on hands-fix are strongly recommended.
V4.5 should be coming soon.
Looking forward to your reviews!
Description
Trying to make 2d and 2.5d more obviously divided on most occasions.
This update was supposed to be released two weeks ago,but I was VERY BUSY preaparing for my graduation exam and some other papers.Sorry for the absence.
If you have any trouble that needs my advice,you can just @me in the discussion area.(Normally I d'like to help everyone.)
However, some questions including weird image production may be due to your own settings(such as negative prompts and upscaler) I'm really tired of somebody giving me a 4 star for this simple reason.
FAQ
Comments (49)
太棒了!
感谢大佬!请问下v4版本的主要特点是什么呢?
赶快把vela2端上来罢,我已经等不及了(急迫)
Vela的更新已经在做了~在探索一条跟cetus系列不同的路线
(主要是朱玉在前 要好好打磨一下 不然嗯骗下载量要被撅的(悲))
大佬,我用4生成的脸更接近二次元,但3.5的光影更立体细节更多,想要4的脸+3.5的光影有办法吗?
很专业的问题啊(其实我也对4的光影很满意的说)
这么说的话我回头做一个小更新好了
或者你可以用MBW的话
把3.5的OUT 层除了03 04 05外的所有层都按1比重搬到4里面
应该就差不多有你想要的效果了()
@Eagelaxis
谢大佬回复,自己接触AI画图还不久....问题有点多,希望大佬可以耐心看
1.刚学的MBW,是把V4的out除了3,4,5外都拉成1,其他in层跟V3的权重甚麽都不用改是吗?我维持了本来的0.5
2.V4的融合比重取了0.25,0.5,0.75,比较了甚至没看出分别....反而是改了输出层的,脸有比较偏V4,光线也有V3.5的感觉,跟我预想有点出入,没想到比重影响那麽少,大佬都取甚麽比例?
3.另外是MBW有主副模型之分吗?我发现用B模型的权重跑XYZ图会出错,选A模型则正常....所以我现在A模型是V3.5,会有分别吗?
就这三个问题,在这边再次感谢大佬回复
@ph15 1.别的都设置成0就可以了(控制变量)维持0.5就跟普通的模型融合0.5比重没区别了
2.我没太看懂你的问题。。
输入层主要影响构图 输出层主要影响画风
一般如果两个模型画风比较像的话 输出层怎么融都看不出来很明显的区别的
想要看比较明显的效果的话 一般比重取0.5起步吧 逐渐拉到0.8以上
3.那个插件跑xyz图有点bug的
我一般就是先出了模型再挨个试
不考虑这个问题()
权重1的意思就是把B的这层完全替代A的这层,0的意思就是完全保留A的这层
后面还要什么不懂的可以跑贴吧问我(novelai吧搜一搜cetus 能搜到的)
@Eagelaxis 明白,谢大佬解答,有问题我会再提
请问有人知道怎么屏蔽特定用户的图片吗?底下有一个人在哪里疯狂的发恶心的图片,就是那个叫Sunbutt的,在好多模型底下都看到他发的图片
在Setting的Hidden User里输入去Sunbutt的名字就好了。
Does anyone know how to block pictures of specific users? There's a guy down there Posting disgusting pics like crazy, that guy named Sunbutt, you can see his pics under a lot of models
I also think it's disgusting,But currently it seems like we can only block he
Sadly we can only block models, not people and their pictures, would be a great addition tho
Reminder: Update your description, it still says' V4 should be coming soon.'
Thanks for the tip XD
佬,可以稍微加一点敏感部位的优化么,就是……那什么……比较逼真的细节要素🤤🤤
那个。。可以用lora做到吧。。之前在评论区里经常能看见的 你可以找一下 有几个老哥挺喜欢发涩图的
但是要融进去大模型的话。。我感觉有点难度 因为得照顾无人物的构图
总不能填个风景出个pussy吧(笑)
回头可以试试
@Eagelaxis 就是因为没有好用的lora……C站搜到的几个相关lora都太差了,要么没啥效果,要么对画面和prompt精确度有严重干扰(大悲)
你小子,欺负老外不会用算盘是吧
LSP
你是真的佬
简直无敌!但是有时候还是会生成多余的手指,有什么好的解决办法吗👀
手这个还是老大难问题。。不过想要直接解决的话可以用controlnet
现在的controlnet版本支持的局部重绘 个人感觉已经很不错了
当然如果你是跟我一样的文生图抽卡守旧派()可以抄一遍我的负面和其他相关设置,我自己用的还算比较舒服
Hey,Can I have your attention qwq
I think the character style of cetus-mix is too heavy, so when I use some other character loras, I can not easyly make my character close to original one.
Can you come up with a version for character?
Sorry for the late response!
If you just want to make the character close to origins,I recommend you using the former series such as V3.5 and V3.Coda and some others are designed for their special styles.XD
By the way,in order to get character loras better displayed,you can also make the lora weights bigger and add some specific prompts to strengthen the character lora.
And I'd put your advice into my consideration on the new updates.Have a nice day!
how is the CODA version different?
Well...Coda is a special series developed from the original Cetus series for its strong contrast between light and shadow,as well as the sort of film texture.
From my own point of view,it can't be described as the formal Cetus versions for that special style,while someone may love it very much and some others don't.XD
@Eagelaxis thanks can you recommend any vaes for the 3.5 or 4.0 version?
음악을 제작하고있는데, 사운드클라우드 앨범 커버로 사용할때에 이 모델을 사용해도 될까요?
물론 가능합니다. Cetus-mix는 다양한 산업 분야의 제작자에게 완전히 무료입니다.안녕히 주무세요(내 시간대에 따라 XD)
amazing, I am going to have sex with the images I'll generate
6
6
6
lmao
6
6
4646
Hello, I'd like to include generated content with this specific model for a game that is set to release on Steam. Is this allowed by the author/is the model trained copyright free?
That's ok. All of my works are free to use for those creators who just wanna use the generated content.(Except some SD model uploading websites.)
:) Sharing the name of your upcoming game on steam is also acceptable.
@Eagelaxis Thanks a lot for the reply, we're still a bit early into discution and research phase but I will make sure to let you know when the time is right :)
Be careful, Steam recently started a campaign against AI assets, and atleast, you would need to tell them that you have licenses to use it or it wasnt trained with copyrighted content.
I this this style has to one of my favourites, appreciate you.
Versatile ASF !!!! Thank You!!!!
Details
Files
cetusMix_v4.safetensors
Mirrors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
SDH3rCwrqE-V4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
Cetus-Mix-v4.safetensors
ceMV4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
cetusMix_v4.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.









