Online multimedia presentations, such as news, need to be constantly updated. Increasing demands are also being made for accessing online multimedia documents from mobile devices such as PDAs. There is an urgent need for a sound but practical formalism that supports automatic adaptation to the change of media content, display environments, and user intention. This paper presents a visual language approach to the layout adaptation of multimedia objects. The underlying theory of our approach is a context-sensitive graph grammar formalism enriched with facilities for spatial representation and specification. The paper focuses on the issues and techniques for size adaptation and style adaptation in response to the change of device requirements and user interactions.