A light pseudoscalar coupled to two photons would be copiously emitted by the core of a supernova. Part of this flux would be converted to $\gamma-$rays by the galactic magnetic field. Measurements on the SN1987A $\gamma-$ray flux by the Gamma-Ray Spectrometer on the Solar Maximum Mission satellite already imply a bound on the coupling $g < 3 \times 10^{-12}$ GeV$^{-1}$. The improved generation of satellite-borne detectors, like EGRET or the project GLAST, could be able to detect a pseudoscalar-to-photon signal from a nearby supernova, for allowed values of $g$.