Aerosol generation during coughing and sneezing has gained major relevance due to the current COVID pandemic. The atomization involved in this process takes place in the complex context of the respiratory system and develops very rapidly. In order to get further insights on the early spray generation, we introduce a simplified model of physiological coughing or sneezing, in the form of a thin liquid layer subject to a rapid (30 m/s) air stream. The setup is simulated using the Volume-Of-Fluid method with octree mesh adaptation, the latter allowing grid sizes small enough to capture the Kolmogorov length scale. The results confirm the trend to an intermediate distribution between a Log-Normal and a Pareto distribution