The concept of differential absorption has been widely used in UV and shortwave remote sensing. This study explores how to extend such concept to the thermal-IR for fast estimation of the total column water vapor (CWV) from clear-sky IR radiances. Using Atmospheric Infrared Sounder (AIRS) radiances as a case study, double difference of radiances at two pairs of pre-selected AIRS channels can be used to suppress the influence of continuum absorption and to highlight contrasts due to weak water vapor line absorptions. To take emission into account, another two AIRS channels are used as surrogates of surface temperature and lapse rate in the lower troposphere. As a result, a three-dimensional look-up table (LUT) can be constructed based on training data sets. Such LUT enables us a fast estimate of CWV directly from the spectral radiances without any a prior information or formal retrieval. The performance of the method is tested using synthetic AIRS radiances based on reanalysis as well as actual sounding profiles. It is also tested against AIRS L2 cloud-cleared radiances and CWV retrievals. The comparisons show that the mean bias of this method is within 7 0.07 cm and the root-mean-square fractional error is about 33%.