I'm curious to know how well the bokeh/shallow depth works on the 7+. By the looks of it they use facial recognition and machine learning to determine what not to blur.

This means objects and figures that aren't faces might not be tuned too well.