An automatic performance apparatus output key data to different kinds of first and second internal or external sound sources and causes the first and second sound sources to synchronously generate tones. Key data is outputted to the first sound source after the lapse of a first delay time data is outputtted to the second sound source after the lapse of a second delay time. When the first sound source, which immediately starts tone generation in response to a tone generation instruction, and the second sound source, which starts tone generation with a delay time from the tone generation instruction, are caused to perform tone generation, a tone generation instruction to the first sound source is delayed from that to the second sound source, thus allowing synchronous musical tone generation between the two sound sources.