Pony diffusion can do stomach bulge well enough out of the box, but these loras improve the concept a lot and give you more control.
Version small: self explanatory
Version nice: large bulge but not absurd
Version pectoral: bulge going through the upper body, as opposed to going away from the body
Version super: absurd stomach bulge size, but away from the body at belly level
Version all: includes datasets for all versions except small
Description
small bulge
FAQ
Comments (6)
That's some fine examples you got, my dude.
1.5 ones were easily my most used LoRa's ever. Guess I'm doing this again.
when i use this lora i get an error message:
F D:\a\_work\1\s\pytorch-directml-plugin\torch_directml\csrc\engine\dml_util.cc:118] Invalid or unsupported data type.Looking at the meta data of the lora... Must have had a pretty interesting dataset, wewewew.
Good lora but I still can't get results I like with it, can you tell me how you made it? Never made one, currently looking up guides but majority are about doing single character and style where you train one concept, this is multiple concepts, so I'm not sure how to label the images for training, what settings to use in kohya and how to control what it will generate if I can only specify a single activation keyword in training. Do you train each bulge type separately and then merge loras or do you actually describe the type of bulge or somehow add multiple activation keywords?
10/10 on those more realistic images