Motivated by the ideas of automatic common-midpoint (CMP) stacking without normal-moveout (NMO) correction, hence NMO stretch, and automatizing the velocity model building, we propose a cross-correlation/cross-coherence-based approach. It is a two-step method where the first step is cross-correlation/cross-coherence of zero-offset traces with all other traces in corresponding CMP gathers. This step removes the NMO effect of different hyperbolic events, resulting in CMP gathers with flat events without any stretching effect. Following this, horizontal summation across different CMP gathers is done, resulting in a velocity-free data-driven production of time-domain stacked seismic section. The second step takes advantage of the cross-correlation lags via data-driven k-means cluster analysis to separate lags corresponding to individual hyperbolic events in the CMP gather into distinct clusters. Different norm fittings to lags within individual clusters are evaluated and the lowest residual one automatically selected, resulting in a velocity and zero-offset two-way traveltime time per cluster. These form a base to build an average velocity model for migration and time-to-depth conversion. We demonstrate the efficiency of the proposed method using synthetic and field shear-wave data acquired in southwestern Sweden.@en