To achieve global climate goals, a greater focus needs to be on the energy-efficient conversion of the existing building stock in industrialized countries. To prioritize the retrofitting scenarios of large stocks of existing buildings, holistic life cycle assessments (LCA) help to consider the environmental impacts in the decision-making. To enable the effortless creation of large building stock information, we propose a methodology to automatically create semantically rich 3D models for calculating the LCA of retrofitting variants. Robustness is achieved by providing flexibility towards input data, e.g., geometric reconstruction based on different point clouds, such as laser scans, drone-based photogrammetry, or derived from Google Maps. Similarly, various image sources are used for the semantic enrichment of windows, such as from hand-held devices or Google Street View. Using a case study, we compare the performance of the geometric reconstruction, test window detection, and calculate first LCA results.