One limitation of Infrared (IR) sounders is that the radiances are affected by cloud scattering and absorption in cloudy skies. Usually, the usage of IR sounder measurements is confined to clear skies. However, it is important to use IR measurements in cloudy regions, as completely clear-sky observations for sounders are typically quite restricted. In an effort to extend the high temporal Geostationary Operational Environmental Satellite (GOES) IR sounding products from clear to cloudy skies, a cloudy sounding retrieval algorithm has been developed and applied to GOES-12 Sounder radiance measurements. This algorithm uses the measurements from all the spectral channels whether they are affected by clouds or not. Comparisons with radiosondes at the Southern Great Plain (SGP) the Atmospheric Radiation Measurement (ARM) site and the ECMWF analysis show the retrievals in thin or low clouds perform as well as those in clear-sky and improve the first guess of Global Forecast System (GFS) forecast. The largest improvements are found in the upper level integrated precipitable water vapor (PW) or PW3; less or even negative improvements in the middle level PW or PW2; and the lower level PW or PW1 is usually improved under thin clouds with the help from the surface observations. In addition, the retrieved cloud parameters are consistent with the false RGB composite images. With the soundings in low clouds and thin clouds, the area without soundings is reduced by more than 50 % in an example case study. The application to a tornadic storm on 24 April 2007 reveals that the GOES cloudy sounding retrievals are more useful at the early stage of storms, revealing more pronounced and extensive large instabilities earlier than clear-sky only retrievals.