The pursuit of perfect balance between realism and anime, a semi-realistic model aimed to achieve beautiful realistic faces with sexy hentai bodies.
This model aims to be flexible with all kinds of NSFW/hentai concepts, so a large portion of the merge will be hentai models.
Description
New secret recipe with added weights, aimed to improve the natural faces and diversity of the model. Maybe more realistic, but the style changes depending on the prompts. Can do a variety of nsfw/sfw images, but tends to draw females and nsfw. If you're having a bad experience with certain prompts, it's probably because of the side effects from the added weights, you should use a different model if so :3
All sample images were created from generation alone without using anything but the model and prompts.
FAQ
Comments (40)
🤗
Welcome back ☺️
AAAAAAAA, THE LONG-AWAITED SEQUEL TO A SMASH HIT!!!! I'M SO EXCITED TO PLAY WITH THIS!
Thanks, can you explain what has been updated in the latest version?
You can read it in the "About this version" on the right side. So much has changed since I last uploaded on civitai, hard to get used to :/
@Bloodsuga really nice to hear about the attempts to improve ethnic diversity, thanks!
And here, I finally had "mastered" V2 only to have V3 come out.... why do you have to be so good at what you do. I cant keep up lol
NANI? But thanks :D
I can't change a model face to a mature woman. Anyone? I need help.
i always write on prompt (milf face) , negative prompt (loli face, child face) , i dunno it'll fix your problem but it works for me
@ilhamamphibia thanks. Btw org indo bro? Kalo iya mau nanya kalo community ini ada ga sih grup nya dari tele, DC, atau FB?
"Denoising strength: 0.72, Mask blur: 1 " means the image was create with inpaint? right?
Yes, most likely.
Clip skip 2 is literally required to get any good images..
That's like saying 20 steps or more is literally required to get any good images..?
did the v3 mixed with AOM3 or AOM2?
All the nsfw elements are thanks to AOM3 :)
涩炸了🥵🥵🥵
This is an amazing model, one of the best i ever used, does it have an inpainting version?
Uh I never understood what's the use of "inpainting version", I just load the regular model I want and use that as the inpainting model in img2img, always worked for me.
@Bloodsuga I dont understand much, but ive seen a lot of models that have there, thanks for replying
@rickmashups To put it simply, I don't know how to make an inpainting version, I always use non-inpainting models for inpainting.
Is there any way to get petite females, with small breast? i always get thicc
Try petite, skinny, small breasts or tiny breasts or medium breasts. I never tried that so the model might not be good at that :(
This might be not a best model for that, try dosmix.
seems like skinny is out. Elephant style is the new thing ;)
May I ask if this model (V3) requires additional use of VAE? The image I generated is very blurry. Which VAE is used specifically?
请问这个模型是不是要额外使用vae，我生成的图片很模糊。具体用哪个vae？
You don't have to use a vae for v3, it's already baked with 840000 vae. Your images are blurry because most likely you did not use upscale. Use x2 upscale with a latent upscaler at 0.6 denoise and see.
I'm using the latest v4 version, and my settings are the same, but why do my images always look too realistic without the blur effect seen in the examples? Thank you.
@Bloodsuga I'm using the latest v4 version, and my settings are the same, but why do my images always look too realistic without the blur effect seen in the examples? Thank you.
不需要使用特定的vae吗？还是需要别的什么描述
Baked vae就是vae已经烤在模型里面的，你可以选择None (无）vae。
Automatic stops working when launching with perfect world model
I guess everything is supposed to look superficial and airbrushed as hell. Is it an Asian thing.
WDYM? Maybe show an example of what you think is superficial and airbrushed as hell and an example of what is not?
No, it's not an "Asian thing". It's literally just an aesthetic preference. Anyone who has the braincells to understand the nature of subjectivity would realize this, rather than picking a fight over it.
I love the bodies of this model but I am having a terrible time with hands. Even with the Hands Beta.
I try to hard force it with Control Net ie OpenPose & Depth but then it makes the images look like crap.
Anyone figure out a nice combo to really use this Dataset?
Im probably going to slam this together in Check Point Merger & test it with some sets that work well with OpenPose & Depth on my next day off. I love this art style.
You can try Good Hands Lora in Prompts and Bad Hand in the negative prompts. Do not set the steps too high. Sometimes it can generate nice hands. Yeah...sometimes. But it is nice
Sorry I never gave a crap about hands, only focused on the face and nsfw capabilities. You can try merge block weights with protogen (the version that gives perfect hands?).
Details
Files
perfectWorld_v3Baked.safetensors
Mirrors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectworld.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors
perfectWorld_v3Baked.safetensors