In the deep mining process, the coupling effect of high stress and high water pressure results in a more complex evolution mechanism of rock mechanical properties. In this work, the degradation effect of granite under complicated conditions is analyzed for analysis and failure characteristics by Nuclear Magnetic Resonance granite initial porosity measurement. A high-temperature triaxial rheological rock system was used to conduct the stress–seepage coupling experiment, introducing energy consumption than energy evolution of granite failure process analysis, and this is combined with the accurate and basic characteristics of the granite building stress–seepage coupling damage constitutive model. The findings reveal that the main pore in granite is the main cause of porosity change, which directly impacts the porosity size and sample strength. Porosity size is associated with secondary pores and main pores, and the porosity is difficult to be affected by the proportion of micropores. The peak strength and peak strain of rock decrease linearly with increasing pore pressure as the decreasing rate gradually increases and increase linearly with increasing confining pressure as the increasing rate gradually decreases. The peak permeability increases linearly with increasing pore pressure but decreases linearly with increasing confining pressure. Rock failure strain energy exhibits an evident confining pressure effect and pore water pressure effect. The peak stress point is the extreme point of elastic energy. After the peak point, the elastic energy rapidly transforms into dissipated energy of rock damage. The energy dissipation ratio increases in the early stage and decreases in the late stage with increasing confining pressure and increases overall with increasing pore pressure. The initial porosity of granite is introduced, and the rock is considered as the solid skeleton and the pore, and the initial nonlinear deformation of rock is employed to build the stress–seepage coupling constitutive model based on the deformation characteristics. The model parameters are obtained by the quadratic logarithm operation method. Compared with the experimental results, the model is found to be of high applicability. The experimental results offer guidance for the analysis of rock deformation characteristics. In the support design of the deep mine roadway, the weakening effect of groundwater on surrounding rock should be fully considered. Effective support of the roadway before the rapid increase of dissipative energy has a significant impact on mitigating mine disasters.