Nowadays, In-Cylinder Pressure Sensors (ICPS) have become a mainstream technology that promises to change the way the engine control is performed. Among all the possible applications, the prediction of raw (engine-out) NOx emissions would allow to eliminate the NOx sensor currently used to manage the after-treatment systems. In the current study, a semi-physical model already existing in literature for the prediction of engine-out nitric ox-ide emissions based on in-cylinder pressure measurement has been improved; in particular, the main focus has been to improve nitric oxide prediction accuracy when injection timing is varied. The main modification introduced in the model lies in taking into account the turbu-lence induced by fuel spray and enhanced by in-cylinder bulk motion. The effectiveness of the new model has been tested with data acquired during an extensive experimental cam-paign during which a 2.0l 4 cylinders Diesel engine, whose after-treatment system allows to fulfil the EU6 legislation limits, has been operated on the overall engine map. It is shown that, comparing measured and estimated NOx on a wide range of engine settings, the im-proved model is quite effective in capturing the effect of injection timing on engine-out NOx emissions: the average error between measured and estimated NOx is reduced of about 10% while the correlation coefficient is increased from 0.86 to 0.97.