A public address system which consists of a longitudinal tubular member extending through an area serviced by the system, with a longitudinal opening and having a plurality of speakers spaced at regular intervals. The audio signals generated from the speakers are emitted through the opening in any direction desired. The speakers are hidden from view within the tubular member and by a cover on the opening which further protects the speakers from ambient contamination. Preferably, the tubular member consists of modular sections, each section including at least one speaker. The speakers may be controlled through a speaker control circuit.