Version 2 has a much better dataset and I also created the new trigger word bgrab.
LoRA for POV belly grab.
Use weight around 0.75 and tags pov, belly_grab, upper body.
Works better with short prompt.