This chapter introduces an action-based neurocomputational model of speech production (ACT) comprising a high-quality articulatory-acoustic speech synthesizer as a front-end module. Some basic features of ACT are the use of sensory information in motor control of speech production including feedforward and feedback processing; a separation of motor planning and motor execution, which can be attained from using the concept of ACTion (the concept of vocal tract action units or articulatory gestures); and the inclusion of principles of self-organization for the buildup of neural mappings for speech production during speech acquisition. The organization of ACT including sensory (i.e., somatosensory and auditory), motor, phonetic, and phonemic neural representations and the gathering of knowledge during learning or training stages of speech acquisition is described. Currently ACT is capable of producing speech items of a ‘model language’ comprising a set of sixty-five CV- and CCV-syllables where C stands for voiced or voiceless plosives, nasals, or the apical lateral.