A voice processing device includes: a sound pickup to receive sounds respectively from a plurality of locations; and circuitry to: identify, from the plurality of locations, a sound source location at which a sound source exists, as a current sound source location; specify at least one location of the plurality of locations other than the current sound source location, from at least one sound source location that has been identified as the sound source location during a past predetermined time period, as a specified source location at which a sound to be enhanced exists; enhance a sound from the current sound source location, and a sound from the specified source location; and output an audio signal including the enhanced sounds.