Hepatitis B virus (HBV) infection is a global public health concern. During chronic infection, the HBV small-surface antigen is expressed in large excess as non-infectious spherical subviral particles (SVPs), which possess strong immunogenicity. To date, attempts at understanding the structure of HBV spherical SVP have been restricted to 12–30 Å with contradictory conclusions regarding its architecture. We have used cryo-electron microscopy (cryo-EM) and 3D image reconstruction to solve the HBV spherical SVP to 6.3 Å. Here, we present an extended protocol on combining AlphaFold2 prediction with a moderate-resolution cryo-EM density map to build a reliable 3D model. This protocol utilizes multiple software packages that are routinely used in the cryo-EM community. The workflow includes 3D model prediction, model evaluation, rigid-body fitting, flexible fitting, real-space refinement, model validation, and model adjustment. Finally, the described protocol can also be applied to high-resolution cryo-EM datasets (2–4 Å).