I seem to recall a company doing this for NPR personalities. It makes a lot sense -- you have one person doing a boatload of speaking in front of a QA controlle

I seem to recall a company doing this for NPR personalities. It makes a lot sense -- you have one person doing a boatload of speaking in front of a QA controlled mic, and you generally have their transcripts. Pretty straightforward dataset for training. Sadly, can't find the podcast that discussed this right now.