typical asian model that does the usual asian face and is less realistic than the old chillout mix. first model on the platform is sus. big community behind it?

为什么出来的都是同一张人脸？这应该是把一张人脸数据精炼到极致了....

作者对男人身体的美一点都不了解，这种肌肉感和比例真假，像塑胶衣一样。女人稍微好点，但脸谱同质化非常严重，像个没灵魂的雕塑。我推荐moonmix或者dreamsharp这类素材比较丰富的模型配人脸lora，多样化才有生气。这个模型真是越来越没意思。

5正式版使用ControlNet会出现边边黑条，别的版本没有，5预览版没有，甚至别的模型都没有，难道只有我遇到吗

Generally good results for lighting and composition but subject's always heavily biased towards a single facial feature. I'm guessing the fine tune images are based on images generated with nwsj's lora. It's unfortunate because the checkpoint's great but is handicapped by having to apply other techniques to create variety in facial features and hairstyle. I'm not saying that nwsj isn't attractive, she is, but a good checkpoint loses its majic if it's only able to yield a single subject.

I’m grateful that the author is sharing this awesome model. It helps me a lot.
He dose not have to share but he DOES!!!!

really amazing model !! beautiful

反应个现象，加载用其他底模训练的lora的时候，大概率出现多个黑痣，似乎算是模型的特点

打下雨或者雨、大雨等有雨出现的时候，头上会出现一把雨伞是啥情况！

大佬，有没试过画美人鱼？尾巴一直画不好，太难了。我试过几十个大模型了，你这几个模型算是其中表达很好的了

试了一下，v5是我目前觉得和我的lora相性最好的模型，gallery发了图，效果真好！

I tried and found v5 performs the best with my Lora, see samples in the gallery!

This model is excellent. Thanks for your hard work !

why is my generated images seems pixelated? i've tried using adetailer and multidiffusion and its still the same. idk if im doing wrong too

很喜欢这个模型, 从V4开始用, 前前后后跑了上万张图了! 但是不知道为什么， 感觉V5的默认脸给我一种更偏向于2.5D的感觉，五官油画质感有点强. 在真实性方面感觉不如V4. 但肢体生成V5还是要远好于V4. 我认为，既然标题中已经提到了“realistic”，如果将来有新版本的话，我希望它能更加贴近写实画风. 再次感谢大佬训练出这么优秀的大模型.

你好，版本V5使用起来对768及以上像素的画面会更接近2D和2.5D的效果，很不真实。以及不知道为什么，我在V5使用以CM为底模训练的人物模型时生成NSFW图片时会出现内衣，但是V4却没有这个问题。

testtesttesttest

downloaded for "research" purpose and ended up doing literal research, wasn't disappointed. This is pretty good at cinematic images in general.

v5 脸很容易崩

When using this on my machine, almost all images are nonsense broken. tried it on Sinkin, same broken results. What am I missing here? Can't generate a single good image.

majicMIX realistic、majicMIX sombre、majicMIX fantasy、majicMIX horror、majicMIX lux这几个有什么区别呢，有没大佬都用过

is this model good for caucasian people or only asian oriented? (:

i cannot for the life of me get good hands to come out of this even using TI negative prompts like badhandv4 and neg-hand. how are people getting good hands from txt2img?

any ideas of inpaint version? =)

新的TG频道连接发一下

u r my fav checkpoint

The moles are getting out of control.

最新版本生成的图片都有一种硬度过高的感觉. 如何改善

great model, but how to make male without facial hair?

why is this model the highest rated, i dont see anything special, just Asian women....

为什么我用V5版本同样参数，生成的图片质感很差呢？像纸片人那样的感觉，谢谢

Which VAE works best with this model?

SameFace strikes again

我下载了MAJIC所有版本，生成的图片质量都很差，所有模型 我都删掉了

我创建了一个服务器我们可以交流然后相互帮助学习lora什么的分享图片什么请大家多多支持https://discord.gg/VFxs3dUsa5

The images I ran on a low memory computer are also very exquisite. Thank you

What is the main difference between V5 and V4?

I'm getting pretty weird hands and fingers on every image I'm creating, how do I fix that?

Here are my negative promts:
EasyNegative, [ :(negative_hand-neg:1.2):15 ], (worst quality, low quality:1.4), bad anatomy, thin arm, long torso, bad hands, extra fingers, mutated fingers, wrong fingers, bad fingers, deformed hand, deformed fingers, missing arms, missing hands, extra hands, mutated torso, partial head, bad face, bad eyebrows, bad eyes, short hands, verybadimagenegative_v1.3, badhandv4, [ :(negative_hand-neg:1.2):15 ], text, nsfw, text, monochrome, nsfw, bad anatomy, bad hands, missing arms, missing hands, extra hands, mutated torso, partial head, bad face, bad eyebrows, bad eyes, short hands, verybadimagenegative_v1.3, badhandv4, text, nsfw, abstract art

打call

v4后无majicMIX