Stabilization and dataset improvements.
You use 1 ~0.8.
The learning model is animefull-final-pruned.Generation model is AbyssOrangeMix2_sfw.