Beyond real or fake: one unified audio LLM for deepfakes
Open turn detector - has the user finished talking?
