>>106378757
see >>106378005
you just need a model that listens but there are limits based on what it's training on.

I like to vary the background so things look more like a scene than a pose. Perhaps you can try that.