A correspondence relationship between virtual logical oscillators, including a master oscillator and at least one slave oscillator, provided for each sound production channel generating a musical tone waveform, and a plurality of physical oscillators for actually generating a waveform and which is associated with the logical oscillators, is stored. Then, according to a process for generating the musical tone waveform, the physical oscillator assigned to the logical oscillator of the sound production channel generating the musical tone is dynamically secured or released with reference to the stored correspondence relationship. Therefore, there is no need to synchronize and playback all waveforms that may possibly be used in the additive synthesis, regardless of whether the waveform is sounded, as is required conventionally. As a result, the waveforms can be generated without needlessly wasting the sound production channel.