From Still Faces To Dynamic Reenactment: Human-Centric Visual Analysis And Generation