Song production in oscine birds is a complex learned behavior that arises from the interaction between the nervous system and peripheral devices. How the acoustical properties of the sound emerge from the delicate and fast control of several muscles in the syrinx and the respiratory system is not completely understood. Here, we recorded electromyographic activity of the syringealis ventralis (vS) muscle and air sac pressure in canaries (Serinus canaria) during singing. We developed a dynamical model with biologically interpretable parameters that translates these motor signals into sound. We show that the vS muscle not only modulates sound frequency but also gates airflow, thereby affecting sound duration and contrasting previous hypothesis in the field. Our results reveal that nonlinearities in the vocal apparatus enable the emergence of complex sounds from the relatively simple motor instructions measured. More broadly, this work advances efforts to achieve interpretable transformations from neural signals to vocal output.