Sound generator for electronic musical instrument

A sound generator for an electronic musical instrument produces sounds from waveform data combinations stored in a memory. Each of the waveform data combinations is a data combination obtained from a waveform of a natural sound by digital data conversion preferably including information condensation. The selection of a waveform data combination is performed by an address generator which produces an address signal for the memory from an information from a keyboard. The information includes a touch information representing key depressing speed and/or strength, or both of the touch information and a pitch information. The loudness level of a generated tone is controlled by a loudness controller in response to the touch information, or to both of the touch information and pitch information.