Recommend:
vae-ft-mse-840000-ema
use highres fix to improve quality.
Source:
using MBW to merge Chillout and Dark Sushi
and also add a little NijiExpressV2
Description
more colorful than v1.0, maybe sometimes too colorful.
FAQ
Comments (64)
伙计,CTUniji0.3fp16这个模型在哪里?
does this model works good for men?
太卷了
sadly to say this, I dont like v2
I think v1 is way more superior.
wonder the point you dislike, color? shadow? or something else?
@Aitasai the coloration is darker, the overall theme of whole set is darker over v1, which is destroying the color pallet, if other people like that its fine, but I dont like it at all over v1.
@Aitasai @Chewtoy I haveto try with 1 some of my generation are in shadow, even when i say they are in sun..
which vae did you use,yours picture is brighter ,i can't figure it out
vae-ft-mse-840000-ema
what upsacler are u using these looks so goood
just latent
Do you use tiled vae/tiled diffusion,CFG Scale fix for generating these images?Also,what are the best resolutions and negatives for this model?I tried generating images using negs from examples but they all got very blurry
I only use highres fix. 20 steps 835*576 , highres fix to 2x and run another 20steps with 0.5 denoising. I havent tried other negative prompts, but I think the examples' negative prompt is ok.
@Aitasai I tried using the examples and also some of my own inspired by them but the images tend to get blurry and pixelated at the end of the generations.I also tried to generate the images you shown into the examples using the exact prompts but the results were far away from what I was supposed to get.
@Corvigel Having the same problem right now, it seems during generation it looks fine then it gradually corrupts as it almost finishes
VAE是否使用了,使用的是哪个?
@Aitasai Can you describe this process better, sorry but I get loss I don't know when you are in text2img or in img2img. Please help us to understand better.
想问一下大佬为啥我出的图脸部细节感觉特别糊呢 完全复制不出展示图的效果
你跑多大分辨率?
有没有装vae
@eric18zhao 我不知道装哪些..
@Aitasai 1024的分辨率
尝试装一下EasyNegative,我之前也复刻不出来另一个模型,也很糊,今天装了配合关键词基本还原了
请问MBW怎么去融合模型,在生成的时候需要配置吗?
okok
I've noticed it's very hard to get a picture where the subject is not in shadow? Even using the terms Key lighting and such puts the person's face in shadow.
I'am very new and a bit lost on this but what's the model you are using??
This is the model.
why use chillout to merge? chillout has license issue and bad skin detail.
bump
I love this model but I'm having trouble getting detailed faces, in a way I don't with other models. The faces are usually noticeably lower detail than the rest of the picture.
Bro can you make a short guide about using mbw, i was experimentin with it and made 25 models to understand what each sliders do but I failed, the quality of generated image was worst than a 50:50 merge in the regular checkpoint tab.
I've got a problem with hands. No matter what I do there's always a problem with hands. Having more than 3 hands or mutated hands etc. Do you have a suggestion for this problem? Great model btw
Use the "bad hands" embeddings and you will be fine. Then include it in the negative prompts and you are good to go.
Look for bad-hands-5 embedding, also with almost all models you have to inpaint the hands a lot to get ones that work. Try ControlNet also
as I see v1 and v2 were just a few days apart and the previews look similar, which model do you recommend using? and what are the key differences between the versions, if possible to know ^^
Absolute masterpiece model !
Are these models strictly for female characters? Is there a way to render male characters as well? Or even both male and female?
I haven't had time to try this model yet but the Gallery below seems to have a good amount of male characters made with it
Results are very blurry, why?
i just cant get faces right... the eyes are always odd hmmm......any tips?
That's when you send it to Inpainting
Use Adetailer
@ZeroKelvin what is this im new to this lol
@LostAtSeaso like i send it, then just mark the eyes right?
@belmont725 Yes, and update prompt for beautiful eyes or something.
Also, make sure you have face restoration turned off when doing anime styles.
@BotHunter @belmont725 just use a detailer ....it does that autmaticall y for u
@BotHunter face restoration usually destroys the style though so you have to be careful
哪个vae好些?
最喜欢的一个大模型,但是手部问题真的很恼火。
what vae did you use?
太厉害了!想问下大佬有国内的兴趣交流群嘛?
wich is the best VAE for this checkpoint?
vae-ft-mse-840000-ema
Wish I could get this without the dark aspect, any light version?
Details
Files
darkSushi25D25D_v20.safetensors
Mirrors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
DarkSushi_v20.safetensors
DarkSushi2.5D_v2.0.safetensors
darkSushi25D25D_v20.safetensors
静静专用末世风2.safetensors
darkSushi25D25D_v20.safetensors
2.5d2.0.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
大寿司2.5darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors
darkSushi25D25D_v20.safetensors



















