In the Gaussian beam (GB) method, initial beam parameters are principal factors influencing the accuracy and computational efficiency of seismic depth imaging. Various optimized beam parameter strategies for Gaussian beam migration (GBM) have been proposed to improve imaging quality as well as computational efficiency, while optimized space-time Gaussian beam schemes for seismic migration have still not been fully investigated. In this paper, an optimized space-time Gaussian beam approach with dynamic parameter control for seismic depth imaging is developed. We first provide an expression for dynamic beam parameter by taking in account the effect of velocity field variation on the beam forming. Based on dynamic beam parameters, the new space-time adaptive Gaussian beam generated by an arbitrary source wavelet is obtained, which can adaptively calculate the beam width to make the seismic beam energy better focused in the central ray neighborhood. Then, the forward wavefield is constructed in two-dimensional (2D) acoustic media by space-time adaptive Gaussian beam for the implementation of migration. Adhering to the framework of conventional space-time Gaussian beam method, we perform the up-going ray tracing from subsurface imaging points to the receiver surface to compute the asymptotic Green function for the construction of the backward wavefield. Numerical experiments demonstrate that the new presented approach has a superior accuracy for seismic depth imaging in both shallow and deep regions compared to the conventional space-time Gaussian beam migration scheme.