A method and unit for composing or decomposing a multimedia signal according to the Musical Instrument Digital Interface (MIDI) protocol. The signal carries events of a first type, which are arranged to carry instructions regarding which predefined patches to use for playback and which predefined notes to play, and events of a second type which carry additional content. The signal is parsed to identify events of the second type and to read the additional content. Coded samples of multimedia content are loaded at an address specified in the additional content. The coded samples are then decoded to provide the samples for playback of the multimedia content. In this manner, vocal song or vocals and other audio type signals are efficiently conveyed utilizing the widely used MIDI protocol.