Figure 1. (b) Composited results with and without expression dynamics optimization. Videos with optimization better follow the original input video's lip contact than those without optimization. The eyes in the other hand follow the intented edited results.