V7 is here. So far so good for me.
ASIAN ALERT!
推荐参数 Recommended Parameters for V7:
Sampler: Euler a, Euler, restart
Steps: 20~40
Hires upscaler: ESRGAN 4x or 4x-UltraSharp or 8x_NMKD-Superscale_150000_G
Hires upscale: 2+
Hires steps: 15+
Hires denoising strength: 0.05~0.5
clip skip 2
如果要修复脸部,请使用after detailer.
If your face comes out badly, use after detailer instead of face restoration.
https://github.com/Bing-su/adetailer
我习惯开启Dynamic Thresholding来更好控制cfg值,1~20都可以尝试一下。
Use Dynmaic Thresholding to control CFG. You can try from 1~20.
https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
如果要添加滤镜效果、噪点,请使用BMAB.
If you want to add noise like me, use BMAB.
v2.5 is here. Note that this is not v7, but an upgrade for v2. V7 may need to address more issues, which I am still trying to do. In the process, I have created a model worth sharing.
This model can restore the light and shadow of the original majicmix v2, but with increased realism while still keeping the basic aesthetic of the face. I have tested other face loras and they work fine. Please use adetailer and put the face lora in it.
请注意,这不是v7,而是对v2的修缮。v7可能需要突破更多问题,这个目前我还在尝试。在这个过程中我融合出一个模型,我认为也值得分享。
本模型可以还原初代majicmix的光影,但是增加写实度,脸部还是保持基本调性的审美,我测试了其他脸部lora都没问题。请使用adetailer然后把脸部lora放在里面使用。
v5对我来说不够好,对你们来说也是的,而且不够写实。所以v6很迫切需要。用Euler画特写,用euler a 画其他的。不要!不要!不要开脸部修复!用adetailer来修脸。我的例图都用了。
v6 is here. v5 is not good as I played for a while and see the results from yours. Therefore a v6 is urgent.
Use Euler for close up, and Euler a for others. Again, don't use facial restoration, use After detailer instead! All my samples used adetailer without lora.
第五版先行版来了,先把赛博永生小姐姐娜乌斯嘉融进来做个例子,也算是公测吧。
5th edtion is coming soon. I've posted a preview version with the face of nwsj.
推荐使用Euler作为采样器。
Use Euler as sampler.
听我一句劝,不要开脸部修复!
Please don't use Face Restoration!
如果要修复脸部,请使用after detailer.
If your face comes out badly, use after detailer instead.
https://github.com/Bing-su/adetailer
我习惯开启Dynamic Thresholding来更好控制cfg值,1~20都可以尝试一下。
Use Dynmaic Thresholding to control CFG. You can try from 1~20.
https://github.com/mcmonkeyprojects/sd-dynamic-thresholding
很抱歉在之前的例图中我使用了分层的lora让大家困惑,也让大家复刻我的例图变得困难。所以新一版的例图我没有使用任何lora。想了解lora分层的可以参考:GitHub - hako-mikan/sd-webui-lora-block-weight
I apologize for using lora block weight in the example images of the previous edition, which confused most of you and made it difficult for you to replicate my examples. Therefore, in the newer editions, I did not use any lora in my showcase. If you would like to learn about lora block weight, please refer to: https://github.com/hako-mikan/sd-webui-lora-block-weight
融合了多种模型,可以生成好看的脸部,也能有效应对暗部处理。远距离脸部需要inpaint以达成最好效果。也可以使用after detailer.
A good looking model, suitable for NSFW and dark scene (because I added noiseoffset). long-range facial detail require inpainting to achieve the best results. You can also use after detailer.
推荐关键词 recommended positive prompts: Best quality, masterpiece, ultra high res, (photorealistic:1.4), 1girl
如果想要更暗的图像 if you want darker picture, add: in the dark, deep shadow, low key, etc.
负面关键词 use ng_deepnegative_v1_75t and badhandv4 in negative prompt
I've used a bug-fixed version of DPM++ 2M Karras, you can check this out: https://civarchive.com/models/35966/dpm-2m-alt-karras-sampler
推荐参数Recommended Parameters:
Sampler: Euler a, Euler, DPM++ 2M Karras (bug-fixed) or DPM++ SDE Karras
Steps: 20~40
Hires upscaler: R-ESRGAN 4x+ or 4x-UltraSharp
Hires upscale: 2
Hires steps: 15
Denoising strength: 0.2~0.5
CFG scale: 6-8
clip skip 2
basic formula:
KanPiroMix + XSMix + ChikMix
关注我的TG频道看更多例图:https://t.me/majic_NSFW
Description
v5是垃圾,v6我还是强调一下写实。
v5 is not satisfactory. v6 is based on v4, more realistic.
FAQ
Comments (138)
才看见,来支持一波
Applause! Finally no longer the same face! Agree with the author that V5 is garbage.
8x_NMKD-Superscale_150000_G 在哪可以下载阿
@Merjic 没看到阿
@a0900998700 https://icedrive.net/s/43GNBihZyi
@tulantuna1 打不开阿
@a0900998700 got any error ?
@tulantuna1 嗯 就是网络打不开
@a0900998700 go here https://upscale.wiki/wiki/Model_Database, click NMKD's Collection
@tulantuna1 您能告诉我选第几个么
@a0900998700 NMKD Superscale > Superscale 8x >
8x_NMKD-Superscale_150000_G.pth
@tulantuna1 https://upscale.wiki/wiki/Model_Database在这里我不知道怎么找 没看到你说的那个NMKD Superscale > Superscale 8x >
8x_NMKD-Superscale_150000_G.pth
ESRGAN4x2哪里可以下载阿 求问
去看插件的描述文件,
@984222978320 我要是能找到 还问啥?
Thank u for sharing. I think that u make the lora nwsj in your model is a wrong decision. People always get aesthetic fatigue. And Your model is wonderful, but the nwsj destroyed v5. Let it be more various or unique. what do u think?
thanks! but I still have to say that nwsj lora is only in the preview not the final.
更这么快?6都出来了
Any tips on getting Adetailer working with deforum?
前幾天才剛下載V5,哭了我
不开修复脸全崩了啊,开启后脸部会轻微模糊,但是结构良好。。。
大佬,我看那个吐司网上有你的模型,但作者好像不是你本人
是我,觉得在线生图效果不错,就留下了
两位大佬
@Merjic 哦哦那就好
@gtbloody 不敢不敢,这才是真大佬 @Merjic
更新了啊?
我为什么屏蔽不掉这个xiodf3130
個人審美V5比V6好看,畫面描述V5比V6好一點
还是v4好看
感谢大佬的模型!是我用过最真实的模型了,从v2一直跟到了v6(更新速度真的是太快了!)有个问题请教一下:在生成人像的时候,比如室内场景,经常会出现比例失调的问题,比如椅子比桌子还大,透视关系不正确的问题,这种问题有什么办法控制吗?再次感谢大佬的付出!
someone please dm me on instagram, I have some doubt regarding generating NSFW image
IG: @dv64bit
用过很多模型,不知道是不是因为1.5模型的弊病,感觉大光圈的景深效果都无法处理的很好,很难做到那种“焦内如刀锋般锐利,焦外如奶油般化开”的效果,这点上midjourney就处理的很好,不知道作者在之后会不会进行优化
我正在疯狂跑MJ摄影,希望可以之后炼个lora出来
@Merjic 期待!
Analog Diffusion这个大模型的景深处理的很好,不知道能不能提取过来
@TangBohu 感觉不太兼容
@HitManLee 我试了效果没问题啊,我发了一张图在这个模型下面,就是模型的第一张图一样的prompt,你自己看看背景虚化的程度吧
@TangBohu 我用了你的prompt可以复现,就是效果还不是很自然,因为这个虚化基本是忽略了深度信息,只对背景进行统一的虚化,所以身体边缘还是那么的清晰,像是在影楼拍照片,后面用幕布作为背景一样。真正的大光圈效果是空间中的某一点最锐利,然后这一点之前或之后的图像逐渐变得模糊,离得越远越模糊,它的虚化是有层次的,您可以对比一下Analog Diffusion和Midjourney或者是大光圈摄影的图像,不知道您是否能改进这个模型。非常感谢您的回复。
Hey,
I have 290k images of faces with captions.
The images are very well diverse.
The captions have been generated through minigpt4.
Can we collaborate
discord: qwerty_qwer#9744
Im trying to train Lora using v6 as base and having a lot of trouble getting good results, with v5 they worked super great. Any tips?
Hi, Merjic. I'm using your model well. There's one thing I want to tell you. When using your model, I found it relatively difficult to apply the characteristics of other lora. The face of the current model appears stronger. I want to show various faces, but I'm sorry that it didn't happen. But it's a good enough model and I want to continue using it. Thanks:)
Thanks! I would suggest you to use afterdetailer and use other lora with it, because I use it all the time even when I was training. But this is an issue I need to fix in the future.
请问可以提供inpainting模型吗?
发现获得高清晰的图片之后,图生图放大,锐度就会降低,皮肤细节就会丢失,不知道该如何解决,如我在下面发的对比图
画面整体偏暗了!
Fantastic model...BUT
It's hard to make different faces than asian look-like with pos and neg prompts, even with Lora's it can't handle it.
This model is very unfriendly to the word sitting
hmmm what is your full prompt?
do u have any inpainting model for majicmix?
会不会出V7呢
不知道为什么,用v4的时候效果很清晰,但是用了v5和v6之后,同样的prompt和设置,结果变得反而很不清晰了,效果出图很差.我用的是mac studio,不知道是否有关.请大佬解答下
感觉如果不指定服装材质的话,很容易出现那种像胶质一样亮晶晶的材质
请问这个模型在upscale的时候,cfg设置在4以上就会使脸上变得特别脏,感觉像是cfg4的设置就感觉像是设置到cfg30了。
使用after detailer插件可以解决脸部生成问题
同感,cfg设置小一点才能出类似的图
majic,C站的希望,昨晚还在抖音看到一个叫麦橘的在直播,不知道是不是你本人
是俺
牛啊 这位作者的模型效果非常好
非常感謝作者提供的這套模型,讓我生成出這第漂亮的圖,我用它製作出了非常精美的模特和各種風格
My Facebook and my AI GirlFriends
這是人最喜歡的模型之一,但它還有一些問題,可能是我沒學會怎麼處理
真的爆杀隔壁chilloutmix
内衣模型只有T字裤,胸衣普遍扁平,除非加大奶量。但样式还是单一。隔壁内衣做的不错,但是隔壁偏正经。
那是爷爷级,没有可比性啦
舊的東西本來就一定會被比下去.跟顯卡不就是一樣嗎.科技和技術一直進步任何東西總有被汰換的一天!
不要拉踩捏
我觉得唯一区别就是chill是暖色调光影majic是冷色调光影,不存在薄纱谁
色调光影可以通过lighting来调节,不存在薄纱谁。
@earachex 真的正经吗
chilloutmix可是一年前的神
CHIL是图腾
请问你主页这么多模型他们有什么区别?不知道该下载哪种
写实、幻想、写实+幻想、氛围感、2.5D、恐怖
@Merjic 谢谢
以前不懂啥叫lora分层,今天刚看了相关资料,又见大佬更新,又又见大佬给了分层数据,啥都不说了,大佬牛逼
大佬,请教一下,V6是原生底膜吗,可以用来做lora的底膜训练吗,泛用性会不会差呀
maji系列做底模容易出女脸,男性角色别用就行
模型效果太炸裂,不知道作者使用了什么魔法来融合多个模型的,期望赐教~~
another asian only model
Another comment ignoring the fact that those non-Asian oriented models are not capable to make a non-stereotypical Asian face, that cannot understand this model become popular not due to the "asian only" part but overall aesthetic value.
Lol i get plenty of other races from this model. Use the prompt box, kid.
@Merjic idk about that, while I agree that this model is picked for aesthetics it's definetly not good for face variety and you can't even make like thai, filipino, etc. It's mostly Chinese / Korean not even asian as a whole oriented model. And faces look very samey with little variety and it can't make average looking face or body.
It's good if you are looking very specific thing but many of those other models can make big variety of good faces so I don't know why you would say that about the other models. Most good ones have bigger variety of faces while this one has very set look.
It is worldwide suitable, not 'Asian-only', bro.
@knoc I agree that other models can make big variety of good faces, but I just can't find the average faces I would enjoy from them, that's why I made this one. I want to make it easier to generate good-looking East Asian (if more specific, my belonging) in a second, no need complex prompt, totally noob focus. I can still change facial identity with other lora through adetailer, not a problem to me.
@st0ic hey..I'm new to this would you like to help me with how do I get non asian look with model? I tried tweaking some prompts but it didnt help.
Thank You
@FooFoo1110 Adetailer has it's own positive and negative prompt. Just use a lora of a face you want in that.
I just use "asian" in the negative prompt. Or "caucasian" in the positive proimpt. Problem solved.
@gr3yh4wk1 There's also an Asian-neg textual inversion here on Civitai.
@nymical I dont use A1111 much, I use NKMD's GUI, which is usually incompatible with textual inversions. But I'll grab that anyway for the odd time I use A111
Ladies and gentlemen, is this a built-in VAE? If not, what kind of VAE should I use?
Set "none" for VAE
None. Yes, Sir! Thank you.
你好,我非常喜欢你的lora模型,也希望能制作类似风格的真人lora,能否提供个联系方式?
V6 感觉很容易出现三只手,然后9:16的经常出现2个头
现阶段配合光影LORA最强真人大模型,没有之一,绝对是出片最真的
哪些光影lora
求大佬贴一下光影lora地址,我们完全不知道要到哪里去找,求求了
Who is artist name on this model actually? Does anyone know her name?
they all look the same
大佬,在使用的时候经常会有一个问题,就是写的是稍微复杂场景一点的词,就会没办法成功,比如”一个女孩和一个男孩,带着一只狗在海边玩“,生成的图片要么没有男孩,要么没有狗,这种的咋办?咋训练能更好
最好用regional prompter
先弄一张满意的背景,图生图在加男孩女孩
这个几把模型是针对单人非全身的效果比较好。你要的还不如直接不用lora
good model
how do I generate a curvy woman (research and anatomy study purposes)
I used curvy figure, slim waist, thick/thicc thigh, but the result is somewhat not what I was expecting. This problem never occurred to me in the previous problem. If any could help, it would be very very much appreciated.
Try more tags like "wide hips" and "hourglass figure", for waist maybe "thin" or "narrow", also add a corset?
@Minterfresh I'm not really familiar with AI terminology, so what is a corset?
@GrandioTestaro i think it is https://en.wikipedia.org/wiki/Corset
@superrazebator327 oh LOL ok I thought it was an AI terminology.
"research and anatomy purpose" LOL
Can I use it for commercial use?
Speak with a lawyer.
good question
I'm not a lawyer, but isn't all AI generated images automatically public domain? Meaning anyone can use them. Not sure about these specific models, but I would lean towards assuming that yes, you likely can.
this is a great model but doesnt seem to work with any TI, anyone having the same issue?
When do friends update? I'm asking for my brothers
Is it just me or it's quite tricky to create a nsfw with this model? 90% of the time it generate swf image, I even try to duplicate one of the nsfw post here and it turns into sfw image
try to put "completely nude" at prompt
Hello, Merjic. I just wanted to ask why can't I see any of your models on the Civitai home page when I'm changing the filtering from 'Year' to 'Month'.
Thx for nice work, this is my most favorite model.
Tutorial for majicMIX realistic easy!
Hi
Its model only for Asian woman ?
works fine with Caucasian's lora
Details
Files
majicmixRealistic_v6.safetensors
Mirrors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicMIX_V6.safetensors
majicMIX realistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
dreamshaper_6BakedVae.safetensors
majicmixRealistic_v6.safetensors
majicMIXrealistic.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
long9.safetensors
majicmixRealistic_v6.safetensors
majicmxr_v6.safetensors
majicMIXrealisticV6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
mjcmxrlstc.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicMIX realistic 麦橘写实_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic.safetensors
majicmix_realv6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicMIX_realistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicMIXrealistic麦橘写实_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
majicmixRealistic_v6.safetensors
long9.safetensors
majicmixRealistic_v6.safetensors
majicmixrealistic.safetensors
Available On (2 platforms)
Same model published on other platforms. May have additional downloads or version variants.

















