The notable developments in pervasive and wireless technology enable us to collect enormous sensor data from each individual. With contextaware technologies, these data can be summarized into context data which support each individual's reflection process of one's own memory and communication process between the individuals. To improve reflection and communication, this paper proposes an automatic cartoon generation method for fun. Cartoon is a suitable medium for the reflection and the communication of one's own memory, especially for the emotional part. By considering the fun when generating cartoons, the advantage of the cartoon can be boosted. For the funnier cartoon, diversity and consistency are considered during the cartoon generation. For the automated generation of diverse and consistent cartoon, context data which represent the user's behavioral and mental status are exploited. From these context information and predefined user profile, the similarity between context and cartoon image is calculated. The cartoon image with high similarity is selected to be merged into cartoon cuts. Selected cartoon cuts are arranged with the constraints for the consistency of cartoon story. To evaluate the diversity and consistency of the proposed method, several operational examples are employed.