The speech of people with Down syndrome presents multiple disorders affecting the different components of language (syntax, semantics, phonology and pragmatics). In particular, prosody is also affected, conditioning their personal development and their social integration. Due to these difficulties, prosody training is fundamental in their speech therapy. One aim of this work is the definition of a video game focused on training some language skills, specifically those ones related with prosody. Other aim is the creation of an analysis system to evaluate the prosodic quality of users’ utterances. The results show that the video game is useful to keep the attention of the players with Down syndrome during all game session. In addition, some statistically significant differences are found between people with Down syndrome and people without intellectual disabilities in the frequency, energy, temporal and spectral domain. The accuracy of identifying a recording as produced by a person with Down syndrome or by a person without intellectual disabilities is up to 95%. Finally, an accuracy of 79.3% is achieved in the task of predicting the prosodic expert evaluation using an automatic classifier trained with the acoustic features extracted from the recordings of people with Down syndrome.