Synthesising Singing

This is a review of some work carried out over the last decades at the Speech Music Hearing Department, KTH, where the analysis-by-synthesis strategy was applied to singing. The origin of the work was a hardware synthesis machine combined with a control program, which was a modified version of a textto-speech conversion system. Two applications are described, one concerning vocal loudness variation, the other concerning coloratura singing. In these applications the synthesis work paved the way for investigations of specific aspects of the singing voice. Also, limitations and advantages of singing synthesis are discussed.