Trained for AbyssOrangeMix2.
Use weight 0.7-0.9.
New version. It more agile but cocks are severly damaged:(
You can also try first version
I saw too many fat cocks today I'll continue work on third version later.
Description
FAQ
Comments (22)
Is there something I should pay attention to when running LORA models, a different (*.yaml) file, perhaps? I can't use them, just returns a lot of non obvious errors. Would love a male version of this, or (even better) a furry one btw
Extension returns this errors?
@FoxDude Loras aren't models, they are add-ons to the model. So you can add this to whatever model you want.
Yeah, tried wit combinations of command line arguments: "--disable-nan-check", "--no-half", and running without "--xformers"
Can't make it work, and if it does, it generates a black image, just like SD 2.0 models would on an SD 1.5 installation
The error generates a lot of lines, but at the end (removed first part of path): "...\stable-diffusion-webui\modules\devices.py", line 152, in test_for_nans
raise NansException(message)
modules.devices.NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check."
Because the error states it might be GPU limitations, others have what looks like the same problem, with a 4090 (Who would be insanse enough to buy that though? lol)
Seen in this thread:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/6923
@Sticks Really? That would be awesome, I just need to figure out how to do that then.
@FoxDude You misunderstand. If using Automatic1111, which it looks like you are, Put lora files into the 'models/loras' folder
Then when you load in your other model, like grapefruit for example, press the sunset button under generate and navigate to the loras tab.
It will get added to your prompts.
@Sticks You're right, I am using automatic1111's UI. Amazing, thanks man!
im jealous
Been having a lot of trouble actually getting this to embed into my images. A version that is less futa would be appreciated, because it seems heavily weighted to that outcome.
Yeah I'd love a version that works better for males
I was looking for a model that does this but its not great in its current form. Changes the output way too much, feels like its overriding whichever checkpoint you use it with.
Not really working for me, its producing oversaturated images weird colors and all that even when using your prompts and model and with the Lora strength toned down, tips?
It's something with your parameters, preview prompts are successfully reproduced
@See Tested again and didn't realize you used SFW instead of NSFW abyss changed it to the SFW one and now it works without making the images oversaturated
In the next version would it be possible to make this compatible with furry characters like anthros/feral e621 style and not just futa content?
It would also be great if it utilized e621 tags and would work well with models like yiffye18, furrye4, and zack3dkinkyV2, as well as some of the more photoreal style furry models, like Stallions (BY Furry Anthro)
Ideally I think this would work best if it exclusively affected character genitals and didn't also force a different style and aesthetic or subject altogether. Maybe it could just replace dicks? I don't know if this is possible but it would be great if it worked this way
The production of the horsecock is great however, it does change the style a lot at higher levels. You might have overbaked it. You could try reducing the training parameters to prevent this. Overall it does produce what you want :p
so creasy
Anyone know how to inpaint with this? IE) can i add a horsecock coming from the side off-frame?
still wait v3
other loras not gens so brutal like this