Model Integrated Computing employs domainspecific modeling languages for the design of Computer Based Systems and automatically generates their implementation. These system models are declarative in nature. However, for complex systems with regular structure, as well as for adaptive systems, a more algorithmic approach is better suited. Generative modeling employs architectural parameters and generator scripts to specify model structure. This paper describes an approach that enables the addition of generative modeling capabilities to any domain-specific modeling language using metamodel composition. The approach is illustrated through an image processing application using the Generic Modeling Environment (GME).