Objective The reservoir of human immunodeficiency virus (HIV) latently infected cells is the major obstacle for eradication of acquired immunodeficiency syndrome (AIDS). Due to the noisy environment and multiple influencing factors in the organism, current dynamical models cannot reach a common understanding of the molecular mechanism of HIV latency. In this work, through a new dynamical structure decomposition, the deterministic part of the equation can be separated from the stochastic noise. Thus, the fixed-point analysis of ordinary differential equation is enough to obtain the different steady states of the system.Methods We established a dynamical model of HIV transcription process by using continuous stochastic differential equations, which simplifies the dimensions of equations needed to describe the system and increases the explorable space of the model. Different states between latency and activation of virus and their relations can be intuitively represented by potential functions and probability distribution functions.Results Based on our model, the influence of different dynamical parameters on stability is quantitatively analyzed, the parameter ranges of the system in bistable and monostable states are obtained respectively. The theoretical basis of this work is verified by comparing the effects of different factors on dynamic bifurcation with the results of biological experiments.Conclusion This paper goes beyond previous discrete stochastic methods, and can quantitatively analyze the dynamic mechanism of HIV transcriptional regulation through ordinary differential equations, which is beneficial to the promotion to deal with the high-dimensional situation, and further study the occurrence and development of AIDS in vivo, so as to guide the design of experiments and search for clinical treatment.