Most AI applications operate in turns: a user submits input, the model processes it, and a response is returned. A delay of a second or two goes unnoticed. Live voice AI does not offer that margin. In a roleplay training session where a psychologist practices a difficult
