Mobile social network is a kind of opportunistic networks composed of a large number of nodes with social characteristic. However the prevalent community-based routing mostly chooses nodes that have optimal social characteristic to transmit messages, without considering the distribution of communities. Hense these algorithms usually result in high consumption of cyber resource and low transmission success when deployed directly in mobile social networks. To tackle these problems, this study proposes a community-based message opportunistic transmission algorithm(CMOT). For intra-community communication, CMOT chooses an optimal path for message transmission from local community to target community by comparing the transmission probability between the nodes. For inter-community messages transmission, CMOT chooses an optimal community path by comparing the community transmission probability. For intra-community in local community, messages are forwarded according to the encounter probability between nodes. The simulation results show that, compared with classical routing algorithms, such as PRoPHET, MaxProp, Spray and Wait, and CMTS, CMOT improves the successful message delivery ratio and evidently reduces network overhead.