It's true that Google can't read the actual content of a video - but they are getting there slowly but surely. First they'll be able to match audio wave forms, then pick out frames as essentially a selection of image sequences.
What Google can do however, is work out when there is a video on page, what size the frame is, how long the content is and how it's encoded/embedded. For additional information - like you say, transcripts, Schema and video sitemaps can be used.
Therefore, to a certain extent, you can get away with putting duplicate video content across your site and showing different meta elements in the video sitemap - but only if you don't host with YouTube. Google can work out, using the aforementioned methods, whether or not YouTube videos are direct duplicates and prevent you from uploading them.
Nonetheless - videos are always "good" content, by virtue of being a rich media type.