We propose a computational framework to predict synchrony of action in online social media. Synchrony is a temporal social network phenomenon in which a large number of users are observed to mimic a certain action over a period of time with sustained participation from early users. Understanding social synchrony can be helpful in identifying suitable time periods of viral marketing. Our method consists of two parts - the learning framework and the evolution framework. In the learning framework, we develop a DBN based representation that includes an understanding of user context to predict the probability of user actions over a set of time slices into the future. In the evolution framework, we evolve the social network and the user models over a set of future time slices to predict social synchrony. Extensive experiments on a large dataset crawled from the popular social media site Digg (comprising ~7M diggs) show that our model yields low error (15.2+4.3%) in predicting user actions during periods with and without synchrony. Comparison with baseline methods indicates that our method shows significant improvement in predicting user actions.