A reproducing apparatus which reproduces speech sounds at a speed faster than a normal reproduction speed. The outputting operation may be performed, making it easier to hear speech sounds when the video searching operation is performed through hearing of the speech sounds in the VTR or the like. The reproduction is adapted to be performed by the other head which is different in azimuth angle during the normal reproduction operation from the double azimuth heads during a given frame reproduction in the 1.5 times speed reproduction with one head of the rotary heads of the normal two heads being provided as a double azimuth head. The noise caused through the reproduction of the adjacent track may be removed.