It isn’t witchcraft though: it’s just damned clever and a major advance in speech analysis compared to what we usually see. It’s not easy, of course. To reach a stage where this is possible requires providing the software with hours of a single person’s voice. This isn’t something that can be used by just anyone, casually.
And it’s still not perfect. It gets around one word in eight wrong. But let it off: It’s still mind-blowing to think that a computer can analyse a person’s voice, translate it and then use knowledge of how the person sounds to recreate what they’ve said as an audio track in a different language.