One could talk and the sound would recorded The AI would see where is the stuttering and could infer in advance where the blocker could happen and thus give warning to users to breath