Exporting the generated audio stems would be slick, too, rather than having to run capture on it.