This DoRA may be useful for some flexibility in your prompts.

Main trained tags:

• scorpion pose, chest stand, back

• handstand

• from side

• faceplant, facing away (a bit harder to trigger)

• stomach (low number of images and rarely works)

• foot archery (an attempt at this can be observed in the gen but... maybe I'll try to train this concept without distractions in the dataset)