In this paper, a two-mass model is used in order to reproduce a particular speaker's voice. One of the aim of this work is to reduce the number of parameters of the model, in order to make its control easier. In the classical two-mass model for example, this was done using the mass-tension factor Q which gives relations between the masses and the tensions [4]. Five characterisation parameters are measurered on a small corpus of vowels as to determine the distance between two glottal flow wave shapes. Then, a genetic algorithm is used as to tune the two-mass model commands for different pitches, in order to extract relations between those commands.