Voice segments to be used by Java applications can be derived from legacy voice segments used by your state table applications or from audio files: in both cases, they need to be imported into the Java space in the voice segment database. VoiceXML applications do not use voice segments: they use simple audio files.