In order to implement a realistic output in a song learning model, we investigated one of the models proposed in the long research quest of Minldin. We were looking for a model with few degrees of freedom, and replicated the RA model proposed in Laje & Mindlin (2002). The model describes three neural populations of RA receiving inputs from HVC: two of which are excitatory and control the syringeal muscles and respiratory muscles (xk and xp resp.); one which stands for inhibitory interneurons. This model is able to produce syllables sounds. We implemented this model in Python language and tried to reproduced several figures of the original paper: a bifurcation diagram and three synthetic syllables.