The prediction of forages feed value is important to estimate the ruminant performances. Near infrared reflectance spectroscopy (NIRS) has been used to predict the in vivo digestibility (OMD) and voluntary intake (VI) of forages using both, forages and faeces samples. Prediction models are usually built using multiple linear regression (MLR), principal component regression (PCR) or partial least square (PLS) regression techniques. Because of the large variability of fresh forage population and source, LOCAL calibration could be well adapted for their application to the prediction of forages feed value. The aim of this communication is to evaluate the potential of NIRS to predict, the feed value of fresh forages using the LOCAL algorithm on faeces samples.