Currently, we create a figure for each frame. We could instead update existing figures to speed up figure generation.