In most video services, users watch only the scenes they are interested in, and look back on the scenes they have watched in the past. In these situations, users typically use a seek bar for seeking scenes of the video. They often have to operate the seek bar many times to get to the desired playback position. In this paper, we aim to support the seeking of specific scenes from video contents. In our preliminary study, we found that users seek scenes depending on when each character appears in the video. Therefore, we designed a system to support seeking scenes using the information of characters. We evaluated the usefulness of our proposed system by comparing it with an existing system. According to our qualitative evaluation, we confirm that our proposed system could ease scene seeking.