The two-mass model is very often presented as a simple but efficient m
odel quite well adapted for the purpose of voiced sounds numerical sim
ulation. It appear however that the description of the flow through th
e glottis is usually oversimplified. To some extent this point could e
xplain the limits of the model. This paper proposes a more precise des
cription of the flow through the glottis. In particular a quasi-steady
moving flow separation point is introduced. A validation of the theor
y under unsteady-flow conditions is presented here. The importance of
the revised flow model for phonation modelling is evaluated and discus
sed using a simple model for the vocal cords based on the well-known t
wo-mass model.