When I first tried, using Google speech to text, I would cut theknown crap out, then play out of my loudspeaker and into a microphone into google. That did a pretty good job of translating, and it was interesting to see its guesses change as context grew. Per paragraph, I would copy from Chrome and paste into a document. But more than a paragraph, Google would often lose the entire text.

Anyway, then I would do paper edits and cut the audio to suit, but not go back to the text for a transcript. And I only did it for somewhat complex constructed stories, where I needed to know all the tape I had. Not for simple interviews. So I didn't produce transcripts.

Now I'm thinking I might actually be able to do transcripts of finished audio, because it accepts MP3s and one pass of editing the text will probably be enough.

Alas, my backlog is more than 100 episodes by now …