Which open tools or libraries are currently the best to extract/remove vocals? Are there some that can even distinguish between speech and singing?