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
FAQ
Comments (22)
Great model but still lacks natural nipples and areolas, it's still cursed by the Flux static breast look.
Static flux breast look, not quite sure what you mean, can you give some examples with/without?
@6tZ Flux has one areola look. Check all the images generated, no matter the shape of the breast, the areola is always the same.
@Learning2025 I've definitely seen many more shapes and sizes, but perhaps not from my chosen preview images. It was not on the list of "diversity" I was looking to represent, but I'll keep it in mind.
I find that it's often tied directly to breast shape and size.
Do you have any models or suggestions for improvements? Datasets or other models.
@6tZ NippleDiffusion - Flux - General LoRA - v1.0 | Flux LoRA | Civitai
They have good variety..
@Learning2025 Thanks for the tip! I'll check it out and include it in future versions if it works well. It'll be a bit further down the line though.
Are we ever going to get away from the dreaded flux chins?
there are lora, in the meantime
are there gonna gonna be pruned version for 6 and 6.1?
You mean a lower precision version? Fp8? I hadn't planned on it but it can always be done if the demand is there.
I have a new version ready though, maybe better to do it there. It depends on which version people prefer.
@6tZ new versions are desperately needed GGUF plsplspls
@side75 Sorry for delay, will get on this sometime soon I hope :)
@side75 Q8 and Q4 are up. Enjoy!
@BobbyKnobby I hope the new quantized versions help you.
Flux is outdated.
Let’s throw it away and work with Qwen.
Oh yeah, so outdated, these pictures are horrible.
That would be a big fat ...wrong
Flux is one of the best models, and version 2 only runs on NASA GPUs.
flux isnt even near at its capacity yet like sd1.5 or sdxl is but looks so good already. flux 2 is too big, qwen takes too long and nobody moved onto it and shit training, only z image maybe if the base is ever released
I tried Qwen online and personally won nothing comparing with FLUX and Chroma (even Chroma Radiance can more than QWEN in some aspects despite being still in development).
IMO Qwen is for that guys who want to generate extremely high-res pictures but in fact it's not a problem for anyone who can into upscalers. So FLUX is still actual. I don't touch FLUX.2 , it doesn't exist for me.
The only advantage of Qwen is a LLM as text encoder. But now we have ZiT with more loyal to system requirements.
BLL SHIT mate u cant see any high class reality with qwen at least rightnow... go go sdxl go go flux...
FLUX is not outdated, but, in my opinion, it did not meet the high expectations. Plastic skin, the same kind of faces with the same kind of emotions. Mannequins. FLUX2 looks promising, but few on this site can test its power. For most, it is not available in good quality, and compromises with GGUF versions do not reveal its capabilities. I think QWEN is very promising. Powerful, flexible, photorealistic to a greater extent than FLUX. The disadvantages of QWEN models are their resource intensity and the need to refine the parts after generation in 4- and 8-step versions. The time spent on generation in regular versions on consumer video cards discourages any desire to use these really powerful and versatile models. The fastest-growing models, in my opinion, are now the Z models. The relative photorealistic nature and lower resource requirements have made these models popular with most fans of neural network creativity. The pace of improvement of the Z-models is impressive, but it has not yet reached its peak. I had the opportunity to test all the mentioned models, including SD1.5, SDXL, PONY and CHROMA. Of these, only SD1.5 is outdated, but it has not lost its relevance for the initial steps in creativity due to the developed ecosystem and minimal resource requirements. PONY and SDXL are the most alive of all the living, they are at the peak of perfection, thanks to a huge ecosystem and continued development by the efforts of enthusiasts who update and improve them to amazing results.