Define your speakers with their own voices on the top left. Write the script with [A]:/[B]: tags — we'll split it into individual turns and generate matching audio per speaker.