This article analyzes the evolution mode of video view counts of nearly 1 000 short videos from “Douyin” platform. According to the statistical law, it is found that there are two time-varying patterns of cumulative short video view counts, namely single gradient and multi-gradient propagation modes. In order to explore the reasons for different propagation modes, this article analyzes the correlation among user’s liking behavior, the emotional tendency of comments and the view counts. It is found that the user's liking behavior will lead to further dissemination of short videos, and the higher the proportion of negative emotional comments is, the larger the view counts of short videos are. Finally, based on the two main spreading ways-passive viewing from recommendation of “Douyin” and active viewing from fans, and taking into account the feature that liking promotes the view counts, this article constructs an online short video liking-spreading dynamics model, which reproduces the evolution patterns of view counts of online short videos well, and the influence of different parameters on the evolution patterns of view counts are explored through the model simulation. In a word, this article reveals the spreading mechanism of online short videos and provides theoretical support for making more effective strategies for information promotion or control.