Several studies have shown a test material dependency on perceived quality. Consequently, methods for evaluating AV quality should take account of influence of content corpus (regarding influence of audio, video and relationship between audio and video) to avoid uncontrolled effects. In this paper, technical, semantic and hedonic dimensions (including different descriptors) were used to characterize a corpus of AV contents. Results have showed significant effects of modality, dynamic and interest descriptors on the viewer AV quality perception. It also seems that verbal sequences and diegesis interact with asynchrony perception. As a result, these descriptors should be considered when selecting test sequences.