An accurate knowledge of the direct solar irradiance at normal incidence (DNI) is required to size solar energy systems, specially those using solar concentration technologies. In the absence of measurements, DNI can be estimated over an arbitrary site with a dedicated satellite model or by using satellite-derived global horizontal irradiance (GHI) and a phenomenological diffuse-direct separation model. This second procedure is error-prone, specially under partial cloudiness conditions and low solar elevation angles. A novel and simple semi-empirical satellite-based model to estimate DNI in moderate albedo regions using a physical clear sky model and a cloudiness index obtained from visible satellite imagery is proposed and assessed. An ergodic assumption in which the images are spatially averaged to better represent the hourly time basis is used. Over the target region (Southeastern South America, SESA), the new DNI satellite model outperforms the alternative strategy at the hourly level, with an average uncertainty of 20% (versus 25%) when compared with three measurement station’s data sets located in center Argentina, southern Brazil and Uruguay.