We present a wide-azimuth beam tomography based on fast beam migration and a method for automatically estimating the rms velocity. This combination allows us to have a direct “input data to velocity model” workflow that does not involve laborious manual user interaction other than quality control (QC). The estimated rms velocity model serves as an initial model for tomography. A 3D residual-moveout (RMO) method enables a very rapid estimation of the depth or time delays along each ray, which represent the direct input to a tomographic update, without the time-consuming steps required for traditional tomography, including preparing the gathers for semblance analysis, semblance picking, and back-projection picks QC. A 2000 km2 velocity model can be updated using 400 CPUs in less than five minutes on a 20 m by 20 m by 20 m velocity-model grid. In addition, beam tomography retains the true azimuthal information. This allows the tomographic update to go beyond the current limitation of limited wide-azimuth velocity updates. Beam tomography allows for faster turnaround time for large 3D seismic projects and, at the same time, increases the accuracy of the velocity model by using wide-azimuth information that is typically unavailable in traditional tomography. In addition to a single-parameter update (one time delay or residual velocity value for each image point) and multiple-parameter update (time delay depends on offset), we now have a wide-azimuth and offset update (time delay depends on offset and azimuth).