This paper describes the design of a speech synthesis for mobile application for Fongbe (Fon Language), a tonal language spoken by nearly 60% of the population of Benin Republic (West Africa). The growing penetration rate of mobile phones in Africa and particularly in Benin Republic makes it important to create a speech synthesis system for this language in the form of a mobile application to help illiterate people. The realization of this application has consisted in the implementation of the Fongbe text analyzer with text cleaning methods, grapheme-phoneme conversion rules, post-lexical rules, etc. A speech corpus of 3007 words was developed and then used to build an HMM footprint for the HTS speech synthesis system. Using HMM speech synthesis system on mobile phone requires some optimizations like selecting the suitable type of spectrum parameters and vocoder. Therefore, the application of 24th order LSP and the single excitation vocoder has been recommended after testing various spectrum parameters. Experiments performed on two mobile phones revealed that the system produces a speech quality with a Mean Opinion Score (MOS) of 2.72 and a response time of less than 3 s.