Sound synthesis by physical models is based on the simulation of the sound production by the instrument. Recently, progress has been made in the modeling of auto-oscillations and in the internal acoustic field of woodwinds. The modeling of the external field, however, its reproduction by electroacoustic devices as well as the control of this type of synthesis, describing the action of the musician, are usually too simple or not even considered. The development of software calculating the sound radiation of woodwinds (far-field) from a precise description of their geometry, has enabled us to compare the directivity of the principal instruments of this family as, for example, the oboe, the clarinet, and the flute. Different diffusing systems of sounds obtained from physical modeling will be presented; each of these systems using different principles for the reproduction of the directivity from the virtual sound recordings. In a previous study [Causse et al., J. Acoust. Soc. Am. 95(A) (1994)], the perceptual difference between a real instrument and such a programmable radiation pattern loudspeaker system simulating a real instrument was investigated. In this talk, results of simple listening tests with these different systems of reproduction will be presented.