Stunned. It's not CGI (is it? Somehow a green screen is used? Or stop-mo-claymation? If so, that can't have been a live event. Can it?)
But it's also not a real device (is it? That's just not possible, is it?).
So how did they do that?
I don't know, but I just know I want one. With an on/off switch for the vocal chip. It seems kinda preachy.