As a novel tool for wave manipulation, higher-order topological insulators can efficiently and robustly localize energy in low-dimensional space insensitive to defects. Nevertheless, the fast design of high-order topological insulators in photonic and phononic systems is still a challenge. In the present work, -symmetric continuum unit cells are explicitly described by the moving morphable voids method, and the concerned properties (topological properties and non-trivial bandgap width) are measured by the band theory and symmetry indicators. Based upon these, a real-time design paradigm utilizing the denoising diffusion probabilistic model (DDPM) is proposed, capable of accurately and fast designing higher-order topological insulators on demand. Using the developed algorithm, it only takes about 0.01 seconds to generate a design in a desktop computer, which improves the design efficiency by 6 to 7 orders of magnitude compared with the traditional inverse design method, and the diversity of generated designs is significantly improved compared with the optimization design results obtained by deep learning based surrogate model. In addition, the generated designs have explicitly described geometric information, so they can be directly integrated with CAD/CAE software, avoiding the post-processing step in implicit description method. This real-time design paradigm based on DDPM can be easily extended to inverse design of multi-physical topological materials and other types of metamaterials.