You can't scroll back and forth through the video to compare different points where he's talking about related themes, you can't text-search to find where or whether he mentioned a particular theme, and when you're reading the transcript, it pauses automatically when you stop to think, so thinking is the default. With the video pausing requires effort, so the default is to not think.

There are things that work better in some sort of video. GUI demos, mechanical movements, some kinds of data visualizations, and facial expressions, for example.