Reproducing dry speech is an important spatial-audio technology as well as rendering desirable 3D-reverberation from a dry audio signal in reverberant space. The key issue is inverse filtering of the transfer function (TF) in the reverberant space. Inverse filtering, however, is formidable and can be instable in a reverberant space, since the TF can often be of the non-minimum phase (all-pass). The inverse filtering is able to render a desirable audio signal at the listening point (target position). However, the variations must be contained in the TF in a practical situation, since the listening position is not always fixed to a single point or the TF itself can be time-variant. We assume that smoothing (instead of localizing) the TF on a target position makes inverse filtering robust. The TF, however, must be smoothed keeping the minimum-phase and all-pass properties respectively. Therefore we propose a method for moving the poles and zeros of the TF symmetrically from the unit circle in the z-plane.