In this paper the realisation of “virtual” wind instruments is analysed, in which the bores are treated as linear, invariant systems, characterised by their impulse responses. The “dry” excitation signal has been obtained from the output of the system, sampled through a microphone placed at the flaring bell of a trumpet, by convolution with a proper inverse filter, obtained from the impulse response measured between the mouthpiece and the recording point. This “dry” excitation signal is the convolved with the impulse response of different wind instruments, thus reproducing the sound of “virtual” instruments. By comparing the results of very similar and much more different instruments, it was possible to conclude that this method makes it possible to finely distinguish between subtle timbric difference among different trumpets. This is due to the fact that the excitation signal remains always exactly the same, whilst when an human performer plays on different instruments, he always compensates for the different response of each instrument, making the differences less audible, and often the execution is strongly modified by this unavoidable feedback. The aim of this work is multiple: the “virtual” instruments can be used in subjective listening tests for the comparison of the “sound quality” of different instruments, for the evaluation of (real or simulated) restoration of ancient instruments, and for preliminary listening tests with newly designed ones, before they are actually built. For validating the repeatability of the technique, a blind subjective listening test has been performed. Three different trumpets and a silver-flute have been analysed, and compared with each other. The statistical analysis of the listening tests confirmed the excellent similarity between the direct acoustic recording and the result of the convolution technique.