Audiovisual active reading is a commonly practised activity in different fields. This creative process consists in describing, criticizing and building an analysis during the reading of a document. More and more tools aim at providing ways to perform part of this activity, built on diferent description models. However, the whole activity is lacking a full support. In this paper, we present our work on a unified model for active reading of audiovisual documents, taking into account each part of this activity. Our approach is based on active reading data and activity modeling so as to propose a pertinent, generic and adaptive model.