Audio is also impossible to edit/correct. You’d have to re-record the whole thing (or embed an audio editing suite in the IDE). And also make sure only one person recorded the comments (for a consistent voice).

AI generated text to speech would be far better than this.