A methodology to generate synthetic bimodal polycrystalline microstructures is presented. This work develops a model that couples a Generative Adversarial Network (GAN) and a classic microstructure generator to create statistically equivalent microstructures that match the surface fraction, texture, grain size, aspect ratio, and misorientation angle distributions of the EBSD data. Bimodal microstructures are composed of coarse grains and ultra-fine grains. The proposed methodology can generate bimodal microstructures obtained via rolling and annealing, severe plastic deformation, powder metallurgy, sintering, or cold-sprayed materials. This work presents the reconstruction of a cold-sprayed Al7050 alloy. Computational research is needed to understand the process-microstructure-property relationship with the goal of characterizing the fatigue life of the coatings. The generated microstructures subjected to tension along three orthogonal directions show different local and average mechanical behavior with changes in the yield stress and hardening.