SBML is the de facto standard to encode and exchange models of biological systems in various formalisms. More than 300 curated SBML models of the repository Biomodels are quantitative. We aim at automatically converting these SBML models into a set of compatible Boolean networks. Here, we present our pipeline SBML2BN, which consists in extracting the model structure from the input SBML model and to retrieve the dynamics of the model by deterministic simulation. The structure and dynamics are then used to synthesise a set of compatible BNs. Finally, the obtained BNs are evaluated. We have tested SBML2BN on 200 models for which we show that it is overall efficient and successful.