update: just reuploaded the model, as the filename had to be changed.
This model will generate images of two men and a woman engaging in group sex,
as the woman is penetrated by multiple penises.
Trigger:
double penetration
penis inserted into her vagina
penis inserted into her anus
frontal → please remove the word frontal from your context as it produces inferior results.
Testing:
This model is not good, but I've produced and seen worse so maybe it can be of use to you.
Personally I'm using it for i2i and combine it with other loas.
The Flux.2 9B version seems to produce far better details when it comes to genitalia.
It was my second attempt, as the first one was undertrained.
This one still is not trained well enough. I won't do another version though.
It's odd but using tags instead of natural language worked better this time.
Using the word frontal in training was a mistake (probably among many others) I used it to differentiate between dp with access from the front, as there is going a Lora for Dp from behind (bent over, on all fours).
About the dataset:
• Suckee: 96 images of different women
• Trained on Realistic professional photography only (couples over 18+ years old)
• Epoch 36 out of 36 Trained on civitAI
• CivitAI Tag caption - not trained on natural language
Description
This model will generate images of two men and a woman engaging in group sex,
as the woman is penetrated by multiple penises.
Trigger:
double penetration
frontal → please remove the word frontal from your context as it produces inferior results.
penis inserted into her vagina
penis inserted into her anus
Testing:
This model is not good, but I've produced and seen worse so maybe it can be of use to you.
Personally I'm using it for i2i and combine it with other loas.
The Flux.2 9B version seems to produce far better details when it comes to genitalia.
It was my second attempt, as the first one was undertrained.
This one still is not trained well enough. I won't do another version though.
It's odd but using tags instead of natural language worked better this time.
Using the word frontal in training was a mistake (probably among many others) I used it to differentiate between dp with access from the front, as there is going a Lora for Dp from behind (bent over, on all fours).
About the dataset:
• Suckee: 96 images of different women
• Trained on Realistic professional photography only (couples over 18+ years old)
• Epoch 36 out of 36 Trained on civitAI
• CivitAI Tag caption - not trained on natural language
FAQ
Comments (8)
X6RB74VSM5YTH0M44CVXQV8YD0.safetensors
cool name brother... 🗿
I've seen it recently in many Loras. I did not see where I could influence its name, as I trained it on Civit-AI. I checked the Post edit options. There does not seem to be an Option to change it. Sorry.
@Adaptalab0r how dare you make us spend 0.0000001 seconds renaming the file you worked so hard to create and then gave to us for free! /s
@rjox LOOOL is just for convenience guys!
EG:
If i share a workflow that uses his amazing loras, and someone read this lora name,
how he is supposed to find the correct lora? there's no way on earth.
A correct name helps to find the original loras and also to archive it properly.
He should just download the file, give it a proper name, then upload it again.
Is really simple
@Agent_Smth
didn't think about taking it down and replacing it. Will do in a few hours. Feels really odd that it is necesarry, but i agree that a well chosen file name can save time.
fixed it. Thanks for pointing it out.
@Adaptalab0r can lose the reference entirely if the metadata doesnt load
@Adaptalab0r AWESOME !