I tried to alternate between a leg spread LoRA that uses western images dataset with low weigh and a leg spread LoRa that's based on anime dataset. the one with anime dataset gives more poses and better on the face