Two text-to-speech synthesis systems were developed for one of the eleven official languages of South Africa, viz. Northern Sotho. A diphone synthesis system, based on extraction of diphones from nonsense words, was constructed. A cluster unit selection synthesis system, based on recordings of sentences containing a selection of most common words in Northern Sotho, was also built. The Festival speech synthesis system was used for both systems. Both of these systems performed well in a subjective evaluation.