I am curious how much video-to-text content represent compared to pure text. I have no idea.