You don't want to do so. Your system will sound horroble. Just imagine, instead of nice, professional voice you will have all sorts of voices at different parts of the system. Besides that, all messages for system announcements are packaged in a single indexed sound file and script knows in/out positions for each message. If you don't like embedded VM engine, just buy external VM system which fits into your requirements. Internal VM system is very basic and intended to work right out of the box