DickTator v1
Use the Lora on its own and do not combine with rebalance nodes.
If you really want to add rebalance to the mix, I would not go any further than 1 as a multiplier but you do you.
Strength: 1.0
Included Concepts
Uncut Erect
Uncut Flaccid
Testicles
Ass
Hairy / Smooth
Anal
Anus
Blowjob
Cum
For the rest, just play along and experiment.
When prompting for penises, it helps to specify the direction of the penis. For instance, to get flaccid, try: "His flaccid penis and testicles can be seen hanging between his legs, pointing down". Should help in 95% of cases.
Disclaimer
It's a first version and I did mess up a small part of the dataset, which means you might, at times, find little watermarks. If that's the case, try and reroll the seed, and it should help. I will have these fixed in the next version.
You may also encounter gens where cum is added when not asking for it. I am not sure why this happens and I believe it might actually be tied to the main model which also loves to add liquid on people when you rebalance the layers. I will hopefully manage to get this eventually fixed.
Regarding cut/circumcised penises: I train these Loras mostly for myself, and my preference is for uncut penises, so I am not planning on adding any cut penises datasets within the training data.
Anyway, hope you enjoy!
Description
V1.0
Release version.
FAQ
Comments (10)
very good
Love seeing male focused loras posted. Thank you.
Great LORA! Especially since you chose to focus on uncut penises. Thank you!
I hope you can get rid of the liquid that shows up all over, somehow. Did you find anyway to avoid this as much as possible?
PS: I tried a solution with flux2klein-edit ... "remove any kind of fluid found on the skin ..." and it worked!
Great job on this, thank you for sharing it!
Impressive work. This LoRA basically turns KREA-2 into a limitless gay porn generator almost bypassing all the internal censorship of the model. It has some slight issues that are manageables in second pass, "white penis" sometimes provide dildo-white appendages to males, "chastity cage" will randomly work. It seems to have no issue managing "regular" and "big" penises but asks for "micro" or "small" and the LoRA seems to totally ignore the keyword. There is also some slight issues when having more than one character, but I think this is KREA-2 being confused with all the "he/his/him" that sometimes provide ambiguous meaning to some sentences.
I have to say Krea 2 really does seem the best model for generating realistic humans. Any chance to share training settings?
After vigorous testing with krea-2 mxfp8 from comfy.org. I think the LORA messes with the framing. It is prone to more close-up framing even though the model without the LORA follows the framing prompt. I found my way through with LORA stacking but just a comment in case you want to acknowledge this for your next version.
Seriously one of the best lora's ever. What amazing quality and just wow. Krea2 + this lora is just WOW! 10 stars!
Love the work you do. Thank you for sharing!