A musical sound generation device including a first memory having a plurality of waveform data, a second memory which stores waveform data read out from the first memory, and a control processor which controls such that, when a sound emission instruction is provided and specified waveform data is in the second memory, the waveform data is read out by the sound source processor, or controls such that, when a sound emission instruction is provided and specified waveform data is not in the second memory, the specified waveform data is transferred from the first memory to the second memory and read out by the sound source processor, in which the control processor controls such that waveform data satisfying a set condition is not subjected to a waveform data change by the transfer and waveform data not satisfying the set condition is subjected to the waveform data change by the transfer.