This is a cool idea:
You can listen to many instances of random English words in context via this interface. I confess that I was not certain about the pronunciation of the name of the country Tanzania (which we discussed a lot in our first coffee hour, given that @rgriscom and @Andrew_Harvey have worked so extensively there!). There are over 1000 instances of the word retrievable via Youglish.
Wikipedia gives the pronunciation as
But a footnote adds:
This approximates the Kiswahili pronunciation. However, /tænˈzeɪniə/ is also heard in English.
But to judge from the content on Youglish, /tænˈzeɪniə/ is far and few between — in fact, I haven’t found a single instance.
Obviously, Youglish seems to be targeting English language learners. But I think this kind of “supercut” interface is one that we could really benefit from in documentation. If we have time-aligned texts, it shouldn’t be that hard to create.