To provide a singing synthesizer suited to singing synthesis of rap music.

A phrase data generation unit 100 divides words represented by word data input by a user into a plurality of segments, generates a plurality of pieces of phrase data each representing a phrase of a synthesized singing voice from each segment and stores them in a phrase database 26. A singing synthesis unit 200, upon receiving a reproduction instruction of desired phrase data, reads the phrase data from the phrase database 26 via a display input unit 50 consisting of a touch panel and the like, and generates a synthesized singing voice using the read phrase data.