The 3D models of buildings and their roofs are very common in today’s architectural planing and analysing tasks. There are three main groups of software for 3D roof and building modeling: manual, semi-automatic and automatic.This study presents a comprehensive review and evaluation of the semi-automatic stand-alone module, CityGRID Shaper, integrated within the CityGRID software framework. A notable advantage of this module is its additional support while modeling through aerial images aligned with the point cloud.The primary focus of this work revolves around functionality and accuracy concerning roof recognition and its 3D model creation, utilizing CityGRID Shaper. The work begins with an overview of diverse software and methodologies employed in 3D roof and building modeling. It proceeds to introduction of the UVM SYSTEMS GmbH and provides a detailed description of CityGRID Shaper, including the underlying principles of the OPALS system for automated roof detection, plane segmentation, and α-shape calculations. Furthermore, CityGRID Shaper’s methodology for generating intersection lines and baselinesis described.Theoretical background on different LODs and roof shapes is presented, outliningrelevant information for this work. Additionally, the required and optional input data sources for CityGRID Shaper are discussed, followed by a presentation of the input data utilized in this work, including details on the location and data collection procedures for two CityGRID Shaper projects.The results of 3D roof modeling demonstrate the capability of the CityGRID Shaper module in recognition of straight roof planes, their segmentation and further 3D modeling, although it exhibits limitations in segmentation of rounded roof shapes. Moreover, the smart-preview function streamlines the rapid modeling of most roof shapes with minimal manual intervention, although it does not extend to flat and shed roof configurations, which must be modeled manually.The evaluation methodology encompasses several phases to address each examined parameter independently. This includes a detailed examination of individual parameters within Realtime Calculation mode, an investigation into the influence of area type on modeling capabilities, and an examination of the impact of different LOD on modeling speed. Finally, a quality assessment of the resulting roof models is provided, based on distances calculations between modeled roofs and the original point cloud data.Furthermore, insights into modeling experiences with CityGRID Shaper are shared, highlighting significant observations for this work. Recommendations for potential future enhancements to CityGRID Shaper conclude the study, offering possible further development and refinement.