Video-to-video synthesis, using machine learning to do less than four hours from beginning to end that weird white thing around it appears human, and otherwise appears inhuman.