V2 Notes: Should be better with photoreal looks. TL:DR Better at realism, worse at concepts. Haven't done too much testing yet, but probably best used in conjunction with other NSFW LoRas to make anatomy details more consistent.
Should be able to generate vore-bellies.
Found to work best with a combination of the following words:
big/huge/gigantic/large/swollen/inflated/distended belly
Description
Refined the dataset, changed training parameters. Better at realism, worse at concepts