In this work, we introduce the windowed generalized phase-shifting algorithms (WG-PSAs) using static and dynamic weighting functions/widows. These algorithms are derived from a weighted least square fitted to the monochromatic temporal fringe, thereby, the selection of the window plays an important role due to the fact that it shall reduce the influence of those intensities jeopardizing the phase estimation. In order to make the best selection, we propose to employ an adaptive/dynamic window which has the ability to detect the fringe patterns that jeopardize the phase retrieval. This window is computed iteratively by analyzing the error between the measured and fitted intensities. Furthermore, we provide the analysis of our scheme using the frequency transfer function (FTF) formalism for phaseshifting algorithms. Finally, we executed numerical experiments with synthetic data in which we compare the performance of the dynamic window versus several static ones from the state-of-the-art; although our scheme is more computationally expensive due to the iterative procedure, it works better than the traditional generalized PSAs with a window included or not.