Many particulate composites show heterogenous mesoscopic structural behaviour due to a random distribution of irregular shaped particles with different sizes ranging from micro to macro scale. Accurate mesoscale models are required to represent actual geometric properties of such a composite with realistic features. In this paper, a cost-effective simulation framework is presented, which can virtually generate three-dimensional mesostructure models composed of irregulate particles with real characteristics. A MATLAB code is developed based on the Voronoi tessellation method and the splining technique. The hybrid Voronoi-spline algorithm can be programmed to control the shape and size of particles based on the experimental data. The examples demonstrated the proper adjustability of the models in accordance with the sieve analysis and shape parameters such as sphericity and roundness. The generated model as a triangulated geometry surface can be directly exported into computational analysis of materials, which uses other numerical techniques including finite element method. (C) 2019 Elsevier B.V. All rights reserved.