A nude/NSFW capable model that focuses on images with a female primary subject.
It is heavily biased towards nude images.
Recommended Settings
Sampler is dpmpp_2m (DPM++ 2M) and the scheduler is beta
No VAE or CLIP is baked in. Use separate sources for those.
Each sample/preview image contains the used workflow. Here's a quick article with simpler more beginner-friendly workflow. This is a recommended starting point.
FP8 Download
For some reason, when you upload multiple formats (BF16 + FP8) on one model, which they now support, they hide the FP8 under the BF16 model group in the side menu...
Early Access is enabled to support the development of new versions and models.
Description
Improved general stability and lots of male images trained and merged in.
The positive thing here is how stable it is for both female and males with this version.
FAQ
Comments (16)
Q_4KS Gives me error in Forge RuntimeError: mat1 and mat2 shapes cannot be multiplied (4096x64 and 256x768) This is the only model I have had errors with.
Which program/tool are you using? Gotta make sure it supports GGUF and that you're loading it in the correct way for the program. It may only work in ComfyUI unfortunately. Forge is weird.
I tried only in forge, Have not had any problems with other GGUF models.
@jsoncy1975 Yeah, Forge issue, not sure how to fix.
I've tried converting the models to gguf inside Forge, but the tools don't work for me, so I'm having to use external tools, and Forge doesn't seem to play nice with them.
yup same problem. I guess only good for comfy
Is it possible to add a trained lora to this? If so, how? :D
Yes, this works fine. Do it the way you do it normally.
If you're using ComfyUI, use a LoRA loader.
If you're using Forge, switch to ComfyUI :D
@6tZ Where can I find more info on "lora loader"?
@MatrisBonusPuer Here's a link to an article:
https://comfyanonymous.github.io/ComfyUI_examples/lora/
Note: This is only for ComfyUI.
You simply route your model through a new node, which loads the LoRA model on top, and then outputs the "updated" model weights. And this is what you connect forward in the chain. You may also need to route your CLIP pin in the same way via the same node.
@6tZ Thanks, works great, It was me :)
i"am sorry
Only the SDXL and the best in this field and don't tell me the opposite? gives a single photo with its data as proof. with better quality???
i"am sorry
Only the SDXL and the best in this field and don't tell me the opposite? gives a single photo with its data as proof. with better quality???
Flux seems to be too big and costly to train, and they appear to have salted the dataset against this kind of use? Sad, but also accounting for its slowness, SDXL is probably the better choice in the long term.
I agree.
SDXL or Pony are the best ones for NSFW.
Illustrious will catch up in with photorealism eventually.
Is there any way to get just a topless generation? My mouth is getting filled up with pussys lol
My apologies, lol!
It's not super controllable.
What you could do is try to prompt for specific clothes for bottom.
Alternatively, you could merge this model with another model, and see if they dress more appropriately :D