Gay sex lora trained against Sulphur 2, and tested primarily with Sulphur text to video workflows, and 10 Eros for image to video workflows. I'm hoping this LoRA can help with gay sex scenes of all varieties since Sulphur seems to have most of this covered.
You can use my LTX2-3 penis lora or underboner loras with Sulfur 2 based checkpoints, but I'm working on training those specifically with Sulphur 2 soon. I used LTX Director custom node and Prompt Relay for some of these showcase videos. Workflows embedded with most of the MP4 files.
Version 2 trained for 8250 total steps, virtually the same dataset as version 1 which was trained with approximately 3000 steps using a video dataset.
Description
Version 2 for Sulphur 2 checkpoint where I've trained it with twice as many steps more than version 1
FAQ
Comments (16)
I can't see the download link for v2??
fixed. thanks!
Very good 😍, thank you for Version 2 🔥🔥
I hope to make a wan2.2 model
I think unfortunately both Suphur and 10Eros are very overbaked towards women so really mess with male anatomy
you should be able to use this with LTX-2.3 base model too
it knows what a penis is, which is promising because you could train custom character loras and have it retain their likeness, voice, and dick from your dataset. Another job AI is taking away from humans - porn star.
meanwhile my girls grow cocks as if their getting paid for it
@anth0nx at least someone is cuz I sure am not given how much time I've invested. You can try lower strength of my lora and if using distilled model and cfg 1, use NAG node so you can caption what you don't want to see in your negative prompt
@coachbate Yes, it seems your models are the most comprehensive across the entire internet.
@evan39102561943 a lot of the OG's seemed to have been banned or left here as far as the gay and male focused stuff. Many left are still using SDXL variants though.
@anth0nx I ran into this also where i wanted a clothed girl to come into a scene and she had bigger dick than my dudes. I think adding regularization dataset to upcoming trainings may help, but not sure if it's worth it since it will be rare use for me at least.
Yes, basically nobody's making new image models.