Towards a Northern Sotho talking head

Abstract

Speech-synchronized facial animation forms an increasingly important aspect of computer animation. The majority of commercial animation products are produced using the English language. Major stakeholders in the industry are the producers of animated movies and the developers of computer games, while the creation of conversational agents for communication in cyberspace and for applications in, for example, language learning is an active field of investigation. It is, therefore, not surprising that most of the commercial facial animation and lip synchronization software caters mainly for English. Northern Sotho, one of the eleven official languages of South Africa, belongs to the so-called Bantu language family and is a resource-scarce (in terms of language resources, technological infrastructure and funding), lesser-studied language of the world. The general question as to whether facial animation tools mainly developed and used for English are appropriate for Northern Sotho speech animation is addressed. More specifically, we investigate what can be achieved with commercially available animation products for English. The paper reports on the process followed, the first results obtained and insights acquired. It is demonstrated that a variety of non-English (Northern Sotho) phonemes can indeed be modelled by tools developed for English by combining multiple different English phonemes and manipulating facial muscles and their actions.