Researchers are creating computer models for voice analysis and synthesis that break the human singing voice into components (such as pitch, duration, and vibrato) that can then be modified to produce a more professional-sounding rendition of the original voice.