A system for transmitting an audio signal in pulse width modulated form on a video signal is disclosed. The audio signal is modulated by varying the time occurance of at least one edge of a pulse as a function of the voltage level of said audio signal at a selected time. The pulse is then added to the video signal during each horizontal blanking interval. Upon receipt of the transmitted signal, the video signal is processed to reform said edge varying pulse. An output pulse is then generated for each varying edge of said pulse, whose pulse width is a function of the time of occurance of said edge with respect to a stable reference edge. This output pulse is then used to generate a voltage signal whose level is determined by the pulse width of said output pulse. The successively generated voltage levels comprise the reformed audio signal.