My first attempt at capturing the dynamics of anal sex between partners with penises. I'm not entirely satisfied with the results, but I decided to share this version because I don't have time to develop it further.
It still needs some practice, so for now I recommend using it with a strength of 1.25-1.75, or with the help of "Krea2 TextFusion Refusal-Reduction LoRA."
As always, the goal was to allow for different angles, positions, and perspectives within a single LoRa. Therefore, in the beta version, not everything works as expected - for example, rear-view positions. Errors may occur - incorrect penis position, two penises, etc.
There's no specific trigger word, but I recommend including the phrase "anal sex," "penis inside anus," "anal penetration," etc.
Have fun testing it out!
Description
First photo - lora with 1.5 strength, second photo - lora with 1.5 strength + Krea2 TextFusion Refusal-Reduction LoRA
FAQ
Comments (12)
Can we go to the water park?
removing watermarks isn't a difficult task
than do a lora yourself smart-ass ^-^/
@EvilFear I do make LoRAs, and sometimes I even publish them.
Recently I gave feedback to someone who made an amazing "THICC" LoRA. I mentioned watermarks. They said yeah good point. Now they released a perfect version.
It isn't hard to be open to feedback.
@pogo so you're using 2 accounts ?otherwise I'm not sure why you're answered me. Of course removing watermarks is not a difficult task, but just posting a very bland sentence like that without any things to say, it's really meh. you do you, but if you thing that's a well insightful structured criticism, it's not.
@EvilFear very bland sentence? Like "do it yourself?"
What is your bar for "well insightful structured criticism?"
Good day sir, I hope this message finds you well. While I was testing your fine product I noticed a peculiar problem. All of the resulting images contained a watermark! Eegads! Surely, this was an error that was honestly overlooked. Might I suggest removing watermarks from the dataset and running it again? Perhaps use cropping or AI blurring to remove the watermarks? Yours truly, Pogo
🫵🤣
@EvilFear and yes Im using my two badass loras as an example. Feel free to use them for immaculate semi-pro modeling realism, peasant.
@EvilFear if it wasn't for the fact that this is not the only lora with watermarks in it, the comment would be different. op even says that they were too lazy. makes you wonder if the dataset is only auto captioned without any coherence and images just thrown into the trainer without resizing for resolutions or ratios. also it is now very obvious what the sources are
@JouzaDouza7135 "auto captioned without any coherence and images just thrown into the trainer without resizing for resolutions or ratios"
woah woah bro friendly fire some of us just want to make models without watermarks over here 😓
work nicely with Krea2 Projector Tuner (BSS) with even less weight.