Uses a robust, event-driven model to reliably select the "Original" audio track over dubbed versions.
