Hidden Markov models (HMMs) have recently become the predominant approach to speech recognition. In this paper, we will elucidate the success of hidden Markov models, using Sphinx as an example of an HMM-based system. We will also outline avenues for future research in hidden Markov models.