Type a message — an LLM answers, neural voice speaks it, and the face moves with the sound (lip sync, brows, subtle head motion, blinks).