Going through my backlog - this generates vintage underwear. Not just high-waist panties, but actual full coverage panties (or "granny panties"). This was generated using hundreds of choice samples to generate the clothing rather than simply a Sears catalog feel. This includes a lot of classic Vanity Fair underwear hence the Lora name.
Using various weights, apply certain art-style Loras and experimenting with models, you can get a wide range of results: from inserting underwear into an otherwise 'anime' image, or recreating a classic lingerie magazine spread. That said, this definitely leans into realistic, and at 1.0 weight, depending on the character model you are using, this can definitely impact look. In particular, a classic 'hair flip' style and slight facial style - but this honestly doesn't come through as much as one would think. This can be mitigated with varying weights, ensuring you have a stable Lora that applies art styles, or focusing on inpainting to get results.
Trigger word: vanityfair
Other triggers: panty_briefs,high-waist_panties, full-cut_briefs,high-waist_panties, underwear, underwear only,covered_bellybutton,
Styles: floral print (there are many styles that will show), lace panties, satin underwear
Art Style: vintage (will give you the classic grainy yellow look)
Negative prompts:
bellybutton, navel - this will prevent the bellybutton from showing through the fabric, a common thing among clothing prompts.
corset, one-piece - this will prevent covering up the midriff. There ARE some corsets used in training if you want to prompt for those specifically
waistband - if you think the waistband looks more like a belt, entering this into the negative prompt seems to help
Other known issues:
You may get derpy eyes. I think this is the model struggling between forcing looking at the camera and most of the images having the example looking away. A "look(ing) away" prompt can help, but also use adetailer and inpainting to fix that.





