It's wild that in a few years automatic translation in videocalls (and podcasts) will be a normal thing.

There is latency in any case so a few tens of miliseconds to generate a translated version won't be a steep price to pay for that. The steep price we pay though is going to be much easier impersonization.