When using the real time speak feature, the original talk begins with a bot that produces responses mainly based on the input