The tech may be possible, but it won't solve anyone's copyright problems. The result would be a "derived work" of the original, irrespective of whether it sounded similar or not.