That's why I'd advise speakers to work on writing the raw transcript
of their own talk, i.e. just the raw text. This is really easy for the
person who spoke, since the two problems above go away, and one can
hope (s)he knows quite well what (s)he wanted to say :) About anybody
else can then easily work on the subtitling itself (splitting the raw
transcript in pieces and syncing on the video), which will be made a lot
less work.