This paper introduces a mechanical model of human vocal system to realize human-like speech under an auditory feedback control. While various ways of vocal sound production have been actively studied so far, we are constructing a phonetic machine having a vocal chord and a vocal tract with an auditory system. By imitating the human learning procedure of voice acquisition, a mechanical construction of a human vocal system is considered to generate natural voice so that it can be advantageously applied to a singing voice production.