I can think of one... podcast cleanup app:

1. Speech to text

2.fix up/edit text with GPT-3

3.text to speech in the original speaker's voice(s), preserving prosody and inflection with Vall-e.

If done with every participant's consent I don't see how it's not legitimate.