Origami has attracted tremendous attention in recent years owing to its capability of inspiring and enabling the design and development of reconfigurable structures and mechanisms for applications in various fields such as robotics and biomedical engineering. The vast majority of origami structures are folded starting from an initial two-dimensional crease pattern. However, in general, the planar configuration of such a crease pattern is in a singular state when the origami starts to fold. Such a singular state results in different motion possibilities of rigid or non-rigid folding. Thus, planar origami patterns cannot act as reliable initial configurations for further kinematic or structural analyses. To avoid the singularities of planar states and achieve reliable structural configurations during folding, we introduce a nonlinear prediction–correction method and present a spatial form-finding algorithm for four-fold origami. In this approach, first, initial nodal displacements are predicted based on the mountain-valley assignments of the given origami pattern, which are applied to vertices to form an initial spatial and defective origami model. Subsequently, corrections of nodal displacements are iteratively performed on the defective model until a satisfactory nonplanar configuration is obtained. Numerical experiments demonstrate the performance of the proposed algorithm in the form-finding of both trivial and non-trivial four-fold origami tessellations. The obtained configurations can be effectively utilized for further kinematic and structural analyses. Additionally, it has been verified that corrected and nonplanar configurations are superior to initial configurations in terms of matrix distribution and structural stiffness.