[ghaff]

A good transcript isn't really that hard. You just need to pay someone--the ML ones just aren't good enough to publish--and do some light editing. I usually do them for my podcasts. (Though the ML transcripts are good enough for skimming content and I use them if I'm just going to be using some excerpts from the audio.)

I actually find that, for talks with slides, I do like having the inset speaker video. What I tend to do is to start off with a full screen talking head intro with mostly the slides/multimedia and a small speaker video.