Hope you enjoy this Lora, and look forward to seeing your posts and reviews.
A Lora to generate "used condoms".
I found that the checkpoint model was not very good at identifying this point, so I made this Lora. It is precisely because of the problem of checkpoint recognition that the generation effect of this Lora is relatively random.
The trigger words are as follows：
She is holding a used condom in her mouth,She is holding a used condom in her right hand,a used condom on her xxx,wearing a round glasses with used condoms, used condoms are hanging on either side of the glasses, And any other usage you want, basically just put "used condom" in your prompt and it will work，several colorful used condoms scattered on her body,condom with semen, If the generation effect is not good, you can add the trigger word "UsedConDom", which is not necessary in most cases.
The strength range is 0.8-1.0, a good starting point would be strength 1.0.
I uploaded the sample video with the prompt words (the video itself was processed in Topaz so the workflow is not included).
Description
FAQ
Comments (8)
I can't seem to generate anything but Asian girls with this.
Let me see your workflow
@alienjacksonlee I used the All In One Ultimate v1.5. He is right. It's always Asian unless you dial down to .5 or so, but then the lora doesn't work. All, double blocks, on top of a character lora, doesn't seem to matter.
@gorokox573628fdfdfd I used the official workflow, the simplest one. I'll post a video with the workflow in the comments section. You can try it and see if you can succeed.
Ok but my man... it seems to have an ULTRA STRONG preference for making Asian women. Not that they aren't a devine gift or anything, because obviously they are, but the point is maybe the used condom lora shouldn't select race for you? The most recent one you added is still vaguely asian.
@gorokox573628fdfdfd Oh, because my dataset is all Asian, I don’t know where to find non-Asian material, which is a pity.
Possible to train lora again with blurred faces to reduce Asian women's face bias, please? I've seen another creater using deface to blur faces and train loras
Will a LoRA weight for wan2.1 be trained using the same training set?