A speech synthesizer having music melody is composed of controller to generate more control signals and an address signal under control of trigger signal corresponding to a synthetic sequence of more basic speech units, memory to store more data corresponding to the synthetic sequence and to sequentially output more data containing correspondent musical scale values, musical scale counter to generate musical scale control signal, and speech/melody generator to input more data and selectively generate synthetic speech, or melody, or mixture of speech and melody under control of control signal and musical scale control signal.