This is my first "decent" lora and it's not perfect. I trained it on about 2000 images I picked, with automatic WD14 captions.
It works best in lower resolutions. In ComfyUI I recommend you to use clip strength between 0 and 0.1 and model strength between 0.1 and 0.6, and in A1111 between 0.1 and 0.2. Having the clip strength above 0.2 sometimes ruins the image.
Description
FAQ
Comments (11)
As the creator of one of the best NSFW models out there (Better Than Words)!... I'm a bit curious, what is intended use case of this LORA?
I had never made a lora before, so I wanted to try training a lora on Civitai.
I use this with OP's Capability XL and like the outputs better than Better Than Words, fwiw.
@Jt20892gkJG8 Nice! Might have to give it a whirl! I do generally find the images of Capability a tad more pleasing than BTW. Can't wait to see what @waterdrinker cooks up next!
i don't really have a use for this but since it's a waterdrinker upload i'm gonna try it anyway
great result ! I'd love to know how you did that oO
Do you have a video explaining or ..?
Can't see the LORA in A1111 after putting it to models/lora. What might be the issue, got the latest pull.
Think you need to change the filename, had same issue in ComfyUI
hi, does it use fp 16? safe tensors variant of sdxl?
im using thi sin code stabilityai/stable-diffusion-xl-base-1.0 .... am i doing something wrong?
Good
Which sampler/scheduler should I use with this?
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.