Same model and same prompt won’t necessarily create the same result, unless I misunderstand how these audio models work.