Trained on images of women trying to hide their assets from prying eyes.
Strongly recommend hires fix.
I'd recommend starting with the trigger caught naked. General non-trained prompting can help solicit different draws, like hunched over or covering.
caught by surprise can help improve the likelihood of the associated expression, especially with adetailer.
This version works the best with high-context subjects (like LoRA's), so if you are doing a generic person, it might help to add more details about them.
Shouldn't need meaningful negative prompting, but disfigured can help reduce the occasional 3-arm image.
Description
FAQ
Comments (11)
Now I need "Caught lacking" lora.
lacking?
@CoCatgirl I'm not certain, but I Googled it & found https://www.reddit.com/r/EnglishLearning/comments/14n6n03/what_does_catch_someone_lacking_mean/
It would be more correct to call lora embarrassed, covering her breasts and crotch. Although depending on the weight it is not always embarrassing. Interacts well with other lora characters and with several girls.
Thanks for the review and sharing your gens. Agreed that the current version has a hard time capturing all of the poses (there are a few other ones, but they come out rarely) and especially some of the subtle posture and face context cues, but the "high upside" versions that were better at those draws aren't quite stable enough.
Planning on doing another pass at some point to improve those. Having these inputs from y'all is really helpful first to see what is and isn't coming through for cross validation on that next round, so I appreciate having a pile of them!
@fansay Also, don't know if they are your speed, but would love to get your inputs on my other models as well if you're interested and willing.
An ENF theme! Bravo!!
LoRa name does not match those shown in prompts for your images 😧.
Sorry about that, I'll update them. Thanks for reminding me.
The reason they're different is that Automatic1111 has started using an internal name identifier for the LoRA that retains the id it was assigned during training. I think using the id from the prompt will still work with that platform, but not sure about others.
@thegipper
No problem mate. I thought I'd just let you know. BTW I really love this LoRa, works like a dream 👍.
Please make an SDXL version of this!