In this study, by using the computer vision technology, authors are trying to add sounds that audiences will imagine from silent videos and to achieve a system which can more strongly express the scene of the videos. The proposed system added appropriate sound information based on movement information that had been obtained from the video. Movement information of the video was obtained by using the Watershed algorithm as two dimensional movement vectors. Appropriate sound information was selected from the data base prepared beforehand by analyzing obtained movement information, and added to the silent video according to appropriate timing.