As the amount of recorded video data continually increases, a lot of teams around the world work to propose original methods on automatic video summarization. A particular kind of video is a rush: a rough draft of a movie or a documentary. In this paper, an approach is proposed to summarize the BBC rushes within the scope of the 2008 TREC Video Retrieval Evaluation task. We propose to summarize the videos by applying two successive steps, each dealing with a particular reduction. First, the stacking step focuses on the reduction of the macro redundancy with respect to similar takes of the same scene (rushes). The second step focuses on the time redundancy of the video flow and can be interpreted as an adaptive acceleration.This approach provides good inclusion rates.