Moisture content (MC) is a critical quality indicator for food drying processing, however achieving in-situ non-destructive analyses of dynamic MC of products during processing is still a challenge. This study developed an in-situ indirect measurement method using Terahertz time-domain spectroscopy (THz-TDS) for real-time MC prediction of foods during microwave vacuum drying (MVD). During MVD, THz-TDS continuously sense the dynamic moisture vapour from the desiccator through a polyethene air hose. The obtained THz spectra were processed to calibrate MC loss prediction models using support vector regression, Gaussian process regression and ensemble regression. Then the MC was calculated using moisture loss prediction results. The best real-time MC prediction results for beef and carrot slices achieved R2 of 0.995, RMSE of 0.0162, and RDP of 22. The developed system provides a novel method for drying kinetics research during MVD and expands the applicability of the THz-TDS technique in the food industry.