The great thing about text editors, and to a large extent also paper, is that they give you random access - you can easily see the whole document and make edits at arbitrary points. It's incredibly ergonomic (there is a reason why books overtook scrolls!) and I doubt voice-driven editing can come close. At the same time, audio is easy for creating conversational-style pieces (e.g. podcasts) but a good final product must always come with timestamped transcripts (how else would one grep and/or skip?)