We study how primordial non-Gaussianities affect the clustering of voids at large scales. We derive a formula of the bias of voids induced from the non-Gaussianities by making use of the functional integral method. In a similar way as of haloes, we find that primordial non-Gaussianities can generate scale-dependence in the bias of voids at large scales. In addition, we show that by observing the cross power spectrum of voids and haloes we could check the consistency relation between the non-linearity parameters f_NL and tau_NL. Large voids (high peak objects) would be good targets since the effects of non-Gaussianities are more prominent while the effects of "void-in-cloud" are less significant.