The volume of video data has increased dramatically in recent years, necessitating the development of efficient browsing, indexing, retrieval, and sharing technologies. Most of the time, users watch brief or summarized videos. With the help of the shortened video, users may grasp the subject matter rapidly. The most important scenes in a video are recognized as key-frames and their overall context is preserved. A logical synopsis is then created by combining these key-frames. Since key-frames include a variety of important and interesting information, extracting them is an important step in the video summarizing process. The literature uses a number of different ways. It is less effective and performs poorly. The goal of this research study was to use the cosine similarity metric to obtain key-frame-based video summarization. Five sample videos from SumMe dataset were used for experimental analysis, and the summarized videos are evaluated by compression ratio and F1-score.