Building realistic facies models helps to improve reservoir description and performance. The proposed methodology is designed as an extensible framework for the integration of geological knowledge by combining the advantages of two well-tried mathematical technologies: (1) the discrete smooth interpolation, which provides an extensible framework for conditioned and constrained interpolations, and (2) the tau model, which allows the combination of partially redundant data. This presentation shows how sedimentological interpretations and sequence stratigraphy principles can be integrated in this framework to build geologically realistic facies realizations. The final facies model may be refined by accounting for diagenesis effects, which can be superimposed onto depositional facies thanks to our solution.