Please note that I wanted to create a LORA. However after training the model all attempts at extraction of the content into LORA were unsuccessful. The content was somehow there but at a very poor quality, regardless of the weight of the LORA or base models used.
There are many enemies of the checkpoints so I was hesitant to upload it. (my original creation was correctly transformed into LORA)
But the creations are in my opinion too good, therefore I decided to upload it after all.
I am making a request - if you manage to convert this to LORA -> just do it, you have my full support!
Anyway, this model is good for generating pussies on their own and occasional full generation with a bit better pussies than typical models, however where it shines is INPAINTING. (And it is even without merging into an inpainting model which still can be done, but I deemed it not worth at the time, since I was focusing on creating a LORA).
All the sample content that you see is generated using that model. The split samples were afterwards inpainted.
Did not play much with prompting since the main idea for this model was to use as inpainting, but since it has a potential to generate full bodies (especially if you provide some names or just run for a while).
UPDATE: Safetensors availble!
Description
FAQ
Comments (3)
Screw the "enemies of the checkpoints" - if they don't want this they don't have to download it. I challenge anyone who says "this should be a LoRA" to make it one.
Thanks for the kind words!
I have a newer version which I will upload soon, the checkpoint in my opinion is great, but again the LORA - not as much. I will upload both however.
I'm not an enemy of checkpoints, but I am an enemy of long, overly specific, and unique trigger phrases. If this had a simple trigger like 'pussy' then its contribution could be extracted and merged into other models. As it stands you can only use it as is. If someone likes it, great, but it doesn't make a contribution to the overall development of better and more capable models.











