MPEG-7, which provides a standardized description for a video, allows one to mark out various properties of a video such as the title, author, dates and content to help facilitate the exchange of video information among various applications. This paper discusses the design and features of a video retrieval system, YuXun. It first uses existing video processing tools to generate the MPEG-7 description file for videos and then establishes full text index for the MPEG-7 documents. Therefore, the user can enter a query string in the system's Web interface and retrieve interested video files quickly. At the same time, the system uses a distributed architecture, which has high scalability, to support mass video information retrieval.