Although the synthesis of names is a new topic, a significant amount of work has been done over the last few years on methods for their conversion. As the potential applications in most cases are targeted for public access, very good name pronunciation capabilities are required. We have been developing rule-based approaches for the grapheme-to-phoneme conversion of names (place names, surnames and first names). Our strategy in the development of morphological- and letter-to-sound rules was supported by a database derived from a telephone directory. To provide a good first name pronunciation, a cluster-based approach has been developed to predict the origins of names. This paper reports on our progress and results toward this development.