I was thinking of using AI generated voice to play out some dialog, but adversarial audio could also be interesting - ie some in-audible sound for humans but that gets transcribed by the AI note taker