But you’d never be able to get the timing to work such that the conversational timing stays this good. And if you degrade that, you’ve kinda defeated the point.
You can't if you do it the obvious way -- sound -> stt -> llm -> tts -> sound. But nobody says you have to do it like this.