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
Uploaded an inpainting model so that I can appear on the leaderboard.
What a good way to update models lol.
FAQ
Comments (187)
Showing latest 167 of 187.
nice model !
https://tusiart.com/ https://www.liblib.art/
These two Chinese websites are stealing models from Civitai.
Actually, we uploaded there
In your opinion, moving = stealing?😂
joker🤡
Author can also upload his models to more than one websites.Not stealing.
使用v7，生成的图每张都是纯褐色的，没有任何人物的，这是什么情况？
嘗試不要用 VAE
可能模型有问题，换个模型或者重新下载模型
@10949221297 我的也是这样的，跟宣传图完全不一样
大大怎麼沒有第一版的了？有看過其他人分享的例圖感覺很不錯，可以再次分享嗎？ 💖😍💖
我开啦
@Merjic Ya! 謝謝大大～ 💖😍💖
“修复版”是什么？
Installed AnimateDiff & Prompt Travel and was trying to locate the issue, it seems like this model does not mesh well with the extensions. Would you mind checking it out?
Morpheus1111 seems to do just fine with it. I don't think the model is the problem.
I am sorry that even myself can't remake my front images (due to different settings and versions of webui), yet I uploaded the lightflow of them so that you can check out the workflows and try to remake them. Read the instruction first. Lightflow | SD webUI Extension - workflow restore & save in one click - realv7_front_img | Stable Diffusion Other | Civitai
Thank you. No need to sorry, you already did a great job !
@IamMarkLee 👍
bro your im idol, this is the best model i know
Amazing model. Do you have Discord or email? I'd lie to chat to you about something amazing
能让乳头变长吗，太短了没有魅力
是我脑子出问题了么， 我觉得新版没有2.5好看。。。
the best model so far
为什么用了clip skip = 2之后生成的图非常阴间
我也是 画面人物面部看不成
what BMAB settings do you typically use for the noise? Also having trouble recreating the soft lighting that you have in most of them.
really nice, many thanks.
使用这个模型去衣皮肤油性程度，乳晕乳头大小和颜色以及胸部大小怎么改变，在提示词里加相关词汇似乎没有什么作用？大家有什么好方法吗？
How do you use this model to determine the oiliness of your skin, the size and color of your areola, nipples, and breast size? Adding related words to the prompt words seems to have no effect? Do you have any good methods?
Try using a LoRA like Oiled Skin: https://civitai.com/models/87685?modelVersionId=93329
or Shiny oiled skin 2.0: https://civitai.com/models/17455/shiny-oiled-skin-20-lycorislora
Let me know which one works best, kind of curious.
用你老婆来个图生图
用controlnet和PS来解决，重绘时首先利用controlnet的线条模式拿到人物线稿，然后在PS里修改线稿上的胸部大小，再用修改好的线稿在重绘里配合controlnet就能跑出你想要的胸部大小。乳晕颜色在PS里调色，油腻皮肤在PS里用移除工具去掉人物身上的白色反光点就能轻松解决。
How to generate gif with text using this model?
快更新XL 我将给予咖啡赠送 辛苦！
I am able to reproduce the V2.5 images and I love to experiment with them, but for some reason I cannot reproduce the images you created for V7, which is a bummer. They look great, and I would like to reproduce it to learn more about the model. I installed all the extensions, I even tried to tweak the settings a bit, but to no avail... Is there something I am missing, like some BMAB settings that are not saved in the image exif data? Maybe you have some settings enabled in the settings of A111 that are not displayed in the exif data too?
I look forward to your reply!
Same here, my previous configurations are stuck now...
BMAB is just a filter, shouldn't be a changer. You can post your works so we can figure it out!
@Merjic Sorry for the long wait, I uploaded the results I got. The images are replicated by downloading the image, and putting it in the PNG info and generating it. Then I also tried to tweak the settings a bit as well, but to no avail. https://civitai.com/posts/1191062
@Merjic I really liked the image with the woman spreading her legs and wearing a fishnet, and it is the one that differs from the original the most.
How do I get better full body shoots? Whenever I try, the face gets distorted. Any advise?
Use adetailer and tune the settings a bit. I find using a lower denoising and a little more inpaint mask blur with smaller faces produces better results.
Also may want to crank up the detection threshold a bit if you're making nude pictures. I've had it detect breasts as faces before.
You should install 'ADetailer'
Use ControlNet combined with openpose can easy fix your problem.
Will there be a sdxl version of this model?
dream realistic model！Salute
A great model, salute to you
MajicMix should be renamed to AsianMix, because they all come out asian, and same face, no matter what I do...
majic=me=asian, therefore, majicMIX=AsianMIX
That's not true. if you add american in the first 3 key words you can make an american face
@zhenghao123 I just tried it, it kind of works. A little better if you add Asian-Less2-Neg. Thanks!
do you plan to make an sdxl version of it?
how can i use ADetailer in comfyui? just help me.
https://github.com/ltdrdata/ComfyUI-Impact-Pack/tree/Main has FaceDetailer
This incredibly awesome.
请问使用什么VAE？我无法生成，是蓝色马赛克
似乎很久没有更新模型了，期待新模型
大佬更新XL版本啊
It makes really nice outputs, but I wish it could make ethnicities other than asians :(
I figured out how to make this model generate decent white girls: use EpicPhotoGasm (or similar) as refiner, starting at 0.3-0.5. Then prompt with caucasian and add Asian-Less2
you're pretty much just using EpicPhotoGasm at that point
@insectwedont627 Nope. My workflow is to gen them using only MajicMix, then take the seed and regen as hires using that seed specifically. You should try it: it results in a great blend between the two; and this model is far better at composition, and so the final image will be based on what this model steers it towards at the beginning. The most interesting part of diffusion search happens at the beginning anyway, when using a converging sampler (this is very important).
@FromOffToOn I don't fully understand this, but I cannot wait to give your suggestion a try, thank you for posting this. :)
@pursuit_of_beauty You're welcome. Check out some of the posts on my profile for examples! If you need any help, feel free to DM or comment and I will help you out.
The quality of fashion design this model can produce is way above everything else
非常棒的模型
Been using v7 and liking it! Was wondering what the best ways were to have the results have no pubic hair, as I keep prompting to remove that and the results are always hairy
Planned an XL versione of It? Cause we Need it 您正在计划 xl 版本吗？因为我们需要它
You really did a great job? by the way, do you have a plan to train a model base on sdxl?
lll
nice
very good
When is majicMIX XL
majic_xl快快出现。麻咪麻咪哄。
非常好
Any rough ETA for next version?
nice
NICE
I love your model. However, since I don't have a good GPU, I can't use your model. Don't you have any plans to enable it to run on the civitai generator?
You dont have 4 VRAM? ... use stable diffusion google colab
hey man you seem to know how to use these files, can you tell me what apps i have to download to use this model?
@masterzed I'd like to, but I'm not skilled enough to use Google Collab.
I'll be busy with a lot of work over the next two months, so it won't be possible for me to learn and use it as easily and conveniently as the civitai generator for the time being.
#NFA
Hy! I'm a testnet participant and I recently downloaded and utilized this model for a Blockchain testnet project campaign.
You can try deploy here, its free cz we are in testnet event
https://lagrangedao.org/spaces
https://swanchain.io
#NFA
@zarvan If you don't have vram, you don't need to download any apps.
Just search 'How to run stable diffusion on google colab' on google.
There will be information of using google drive and colab to run stable diffusion.
You only have to download this model and follow the guide that you searched.
什么时候出XL的啊
gotta say DAAYUM !!! v6, v7 are freekin awesome!!!, keep up the good work
Nice
wow
Request
Any chance to make it in Pony?
Have you tried 3010nc in Pony? This model can be considered as the advanced version of Majic.
3010nc is 2D instead of 写实
Great
Excellent
神中神！用上hires fix后真是极度清晰！这个模型的照片感和美感，我甚至觉得比我试过的所有SDXL模型都要好！
best for realistic
请教作者：我在negative prompt那里设置(makeup:1.5)都没什么用。如何有效生成素颜图像？
P
Mm, so sexy
any idea why this isn't available for on site generating? with its 10K positive votes on this model alone, and close to 800k downloads, i just constantly see the best checkpoints, loras etc. in SD,XL,Pony and starting with Flux - not available onsite generation as if theres a reason...
Because on site generations are ugly overall
any example workflow?
cannot reproduction
how do i get a consistent character?
any way to not have pubic hair? like holy hell man
Just put "pubic hair" in negative prompt
cool
Normal
will you ever do an xl version of majicmix? I'm still waiting 🙏
or pony, either is okay hahahah
Nice
Beautiful
ASIAN ALERT! -- I love it! Works very well
What permission should I use? The model is terrible
总是生成多余的腿
probably using too high base resolution, sd 1.5 is most comfortable at 512x512, and maybe 768x768, if you try to do 1024x1024 or higher it'll cause issues like that, you need to use an upscaling script like "High resolution fix" to do a second pass on the initial lower resolution image to get it to a higher resolution without errors.
Hello what size where the examples images generated at before upscale? 512x??? or higher??
ASIAN ALERT! -- I love it! Works very well
Good to create realistic looking East Asian people but fails to create a dark/brown skin person.
nice!
wow
this is so good work boes , is it flux?
🤘🏻😎
anyone know why i cant use this model
what platform you used? this is SD1.5 model
@suihou827 I'm new here, could you teach me.
@suihou827 yes i have used sd 1.5 model
@rbcom OK, can you please describe more what problem you are facing or any error message you have and see what we can help? I used this CP for over a year without any problem.
@suihou827 i think new system is the problem looks like i have to win this cp in bidding first then i can start to generate
@rbcom LOL .......
@Mister_X_peach Yea its pretty easy
@rbcom u have a laptop or pc more than 2GB Vram?
@thatsca197 @thatsca197 @thatsca197 no i dont have graphic card in my pc
ASIAN ALERT!
do people get triggered by asians?
sometime its good
Gorgeous
Damn boi
nice
nice
nice
I wonder if this model can only change clothes, but not other content? For example, hair color?
what's recommended
positive and negative prompt for this?
if possible could you put per version if there's a change in recommended prompts for positive and negative
好的
i download the png ,and load the workflow, run the model but image is blur, any suggest about this?
fixed after update negative prompt
good
i would be very grateful if anyone can tell me what could possibly be wrong
recently downloaded v7 and use it in comfyui, but it can only produce strange, abstract painting-like things.
whether using a simple workflow with some simple prompts, or directly paste the sample pictures (so embeded workflows and parameters go into comfyui), doesn't make the results any better.
i tried to change cfg, steps, sampler (euler, euler_a, dpmpp2m, dpmpp3msde), scheduler (simple, karras, normal, exponential), but none seem to work.
i even checked the hash of safetensor files.
i have 4060ti16g and can generate images with several different models (like wai-illustrious or dreamshaper-xl) without problem. and the comfyui tells me the vram is more than enough.
dear owner, do ever plan to make sdxl version of this model?
A very commendable model. And moderately diverse and stylish, and with a good realistic bias.
<lora:è¢çç¾½-09:0.3>,<lora:skin_tone_slider_v1:-0.5>,<lora:curly_hair_slider_v1:2.1>,<lora:shojovibe_v11:0.4>,
I cant see the name of first lora, can you tell me what is that lora?
8x_NMKD Superscale_150000_G放大模型如何下载啊老师，研究一晚上没研究到地方
hf找找
please make one for illustrious
我今天刚安装了comfyui,这是我下载的第三个模型，但我不知道该把这个模型放在什么文件夹内，我尝试放在ComfyUI\models\checkpoints路径下，但在没有生效，所以谁能帮助我解决这个问题吗？
为什么这个大模型总是生成两个女性甚至更多
Hihi
Details
Files
majicmixRealistic_v7-inpainting.safetensors
Mirrors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicMIXrealisticV7inpainting.safetensors
majicmixRealistic_v7_inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors
majicmixRealistic_v7-inpainting.safetensors