In this article, we propose a conceptual model for video content description. This model is an extension of the EMIR² model proposed for image representation and retrieval. The proposed extensions include the addition of some views such as temporal and event views that are specific to video documents, the extension of the structural view to the temporal structure of video documents, and the extension of the perceptive view to motion descriptors. We have kept the formalism of conceptual graphs for the representation of the semantic content. The various concepts and relations involved can be taken from general and/or domain specific ontologies and completed by lists of instances (individuals). The proposed model has been applied on TREC video 2002 and 2003 corpora that mainly contain TV news and commercials videos.