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
A new and improved version.
Huge thanks to alcaitiff and their Mystic XXX LoRA!
FAQ
Comments (15)
you know, for this to make NSFW Content? well there you go.
I'm getting excellent results with this checkpoint, fantastic work! However it's virtually impossible to hide nipples. Not that I'm complaining though :)
Oh, woe is us :)
Yes, you are right, it's overfit, and nipples are poking through most clothes.
Hopefully we are able to tune it down in the future so that we don't get that error, but still get the good naked bodies.
Hi, just wanted to say thank you! Question: between this and your checkpoint "Fux Capacity" what are the major differences and usage for? Thanks!
You can tell it's AI by looking at the preview pics etc but with SDXL nsfw models the pics looks almost as they are real. Flux have that weird AI look and overall effect on the images. Nice attempt though, good work.
Yes, and all the women look the same. The all have extremely exaggerated cheek lines. Nice work on the nsfw though.
Flux can make incredibly realistic looking images or very plastic looking images, just depends on your generation settings, prompts, LoRAs, etc. SDXL has the benefit of way more community model development but is definitely an inferior model, it does a terrible job with hands, feet, and teeth. It is way worse at character LoRAs, Flux's character LoRAs are so much more accurate and realistic. Just lower your CFGs and experiment with different modifier loras if necessary, there are many that are designed to reduce Flux's negative qualities.
civitaifreezable801607 Generally flux is good yes but it is just sucks you can't use it with 6GB GPU. I can't use it even with Forge.
civitaifreezable801607 mostly, i got no problems with hands or feet anymore, well...at sdxl... in flux on the other hand...its imho much more problems with those
gorathan274 Hands and feet get more messed up as you introduce NSFW concepts. It's like it hasn't seen enough of genitals + hands/feet together...
just downloaded early release and I am eager to make it work. However I get an error on generation try in Forge... : "AssertionError: You do not have VAE state dict!".
I am using your model succesfully in ComfyUI, but I use both tools for different things, so would love to get Forge running too.
I have added both the clip_l.safetensors and the t5xxl_fp16.safetensors files to both the vae and text_encoders folders (just to be sure), and I have selected both files in the VAE/Text Encoder dropdown, next to the checkpoint selector, in the top of Forge. Read online that this should do the trick, but I does not. I still get the error.
You added Clip and T5, but you also need VAE: https://huggingface.co/ffxvs/vae-flux/tree/main
Where could I find the Clip and T5? I got the: AssertionError: You do not have CLIP state dict!
I got it, didn't we could load a T5 and clip at same time. Still learning
@bolo2626 So much to learn! You can also load 3 models at the same time. You can add a clip_g to the formula.