The chatter vibrations in milling are avoided by selecting spindle speeds and depth of cuts from the stability lobes. However, it has been well observed that the structural dynamics of the spindle system change as a function of speed and even feed. As a result, the measurement of spindle structure's Frequency Response Function (FRF) at zero speed does not lead to an accurate prediction of stability lobes, hence the productive cutting conditions cannot be estimated accurately. In this study, an in-process identification of speed dependent FRF of the spindle structure is proposed. The spindle FRF is either measured during the rotation via impact testing, or FRF is identified from the inverse solution of critical stability limits. The tool-holder is decoupled from the spindle FRF using receptance coupling method, which leads to speed and/or feed dependent spindle shaft FRF. The tool tip FRF is then predicted by coupling the Timoshenko beam based tool-holder to spindle. The proposed method is validated in milling tests. It is shown that the chatter free material removal rates can be more accurately predicted when the spindle speed and feed dependent machine tool dynamics are considered.