"how do they keep zooming into the still pictures while the characters are animated using crossfade type of e ffect. Did they seperate the character and the background picture in seperate layers?"

Looks  like layered renders to me,  I usually render .png sequences as that preserves transparency. If you have multi-layered video editing then using the same image sequence at different opacity and slightly out of phase can create the time-blur effect used on some of the headshots in the clip you linked.