In a Karaoke apparatus, a sound synthesizer reproduces an instrumental accompaniment of a song requested by a singer during the course of a vocal performance of the song by the singer. A whole of the instrumental accompaniment is provisionally divided into a plurality of blocks each having a start address and an end address. A detector successively detects start and end addresses of each block with progression of the reproduction of the instrumental accompaniment. A skip command is inputted by a skip switch when a certain block is being reproduced. A controller responds to the detector and the skip switch, and operates when an end address of said certain block is detected for controlling the sound synthesizer to skip an immediately succeeding block to thereby continue to a start address of a further succeeding block during the course of the reproduction of the instrumental accompaniment. Further, a repeat command may be inputted by a repeat switch so as to effect repeat reproduction of a certain block.