We still need a different approach to create this. Consistency is the problem in video, a model should create from start to end instead of different shots.