It is not a direct improvement over v1, rather v2 aims for a softer, rounder, more comic-like style compared to it. It also moves its attention to the whole picture, instead of just focusing on mainly the face and hair. As such, v2 inherits most of the strengths and weaknesses of v1.
This version is less likely to produce random polygons in the background and it should be more consistent across different checkpoints, but is not compatible with as many models as v1.
It does, however, hate covered shoulders even more than v1.
In conclusion, neither v1 or v2 is better or worse than the other.
Second version.
It is not a direct improvement over v1, rather v2 aims for a softer, rounder, more comic-like style compared to it. It also moves its attention to the whole picture, instead of just focusing on mainly the face and hair.
As such, v2 inherits most of the strengths and weaknesses of v1.
This version is less likely to produce random polygons in the background and it should be more consistent across different checkpoints, but is not compatible with as many models as v1.
It does, however, hate covered shoulders even more than v1.
In conclusion, neither v1 or v2 is better or worse than the other.