TED has a application/ld+json block on their pages if you access their URLs with /transcript appended (e.g. https://www.ted.com/talks/gus_worland_is_someone_you_love_suffering_in_silence_here_s_what_to_do/transcript ) so the summarizer could extract the text from there.
Judging by a superficial test, when given a /transcript URL, the summarizer WILL summarize the transcript. So maybe all that's needed is some kind of search & replace regexp where, if a TED video is referenced, it insteads fetches the transcript page (cause otherwise it will only summarize the video description, which usually leads to just some fluffy summary including viewer counts and whatever else it finds of the main video page)
