Lustly is now periwinkl.ai
Check out our completely new, amazing model based on Z-Image. Check out the examples: https://civarchive.com/user/periwinkl/images
======= ARCHIVED INFO BELOW =======
We are thrilled to be introducing our first uncensored Flux model:
Flux Lustly.ai Uncensored v1: Full Frontal Nudity!
You've seen female nudity in Flux; now, we're evening the playing field by introducing male nudity as well.
Performance:
Reliable on Flux dev
Promising on schnell model, though slightly less stable. High-quality results are still very achievable.
Testing:
We tested on full dev and schnell models using diffusers.
On-site CivitAi generator seems to work fine (https://civarchive.com/images/35833806)
Community Call: We haven't tested in other environments like Forge or ComfyUI, nor on quantized models. Let us know if you achieve good results!
For more details on diffusers inference see https://huggingface.co/lustlyai/Flux_Lustly.ai_Uncensored_nsfw_v1
Roadmap
Looking ahead, we're excited to develop further iterations of our open-source base model, expanding into:
More diverse poses
More, ehm, dynamic interactions between individuals
All sorts of kinks
Join the Movement
Try this model out at Lustly.ai
Description
FAQ
Comments (10)
So this can drawn naked bodies well, but still does not uderstand sexual positions? Right?
Yup, this iteration is focused on basic frontal nudity with no sexual interaction. Stay tuned for further updates as we add more capabilities to the model
@lustlyai Nice. You are probably aware of this, but
https://www.reddit.com/r/StableDiffusion/comments/1fuukwz/fluxdevdedistill_an_undistilled_version_of_flux/
Apparently with an undistileed version of FLux fine tuning works better. The guy is quite faat at asnwering, if you have any questions.
@barlogspam176 yeah, this was attempted before as well https://huggingface.co/ashen0209/Flux-Dev2Pro
@lustlyai Ah damn. So apparently the results is not that impressive.
Have you tried funtuning one of the two? Are the results better or worse?
I have noidea how much it costs you to finetune it, so maybe my question is too arrogant.
@barlogspam176 training with de-distilled and Dev2Pro tends to produce better results (see my de-distilled finetune for an example). That being said, if using de-distilled specifically you'll need to use CFG at inference to get good results, meaning that you'll have to use a de-distilled checkpoint which seems to have spotty compatibility with distilled resources.
@barlogspam176 quality sucks when applied to distilled models. The community is still figuring out how to work with flux, for now there's no silver bullet
@lustlyai damn
Wonderful!!! Fantastic results 😊
You should try to create a flux checkpoint like this lora.
Details
Files
flux_lustly-ai_v1.safetensors
Mirrors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
Flux Lustly AI v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
flux_lustly-ai_v1.safetensors
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.