We do it with a few speech to text APIs.
We've only tested with English so far but it should be able to handle a bunch: Arabic, English, Spanish, French, Portuguese, Japanese, and Mandarin.