Most shoe manufacturing processes are not yet automated; it puts restrictions on increasing productivity. Among them, adhesive application processes particularly are holding the most workers and working hours. In addition, the working environment is very poor due to the toxicity of adhesive agents. In the case of automating an adhesive application process by using a robot, robot teaching by playback is difficult to produce high productivity because the kinds of shoes to be taught mount up to several thousands. To cope with it, it is necessary to generate robot working paths automatically according to the kind, the size, or the right and the left of shoes, and also to teach the generated paths to a robot automatically. This paper presents a method to generate three-dimensional robot working paths off-line based on CAD data in an automatic adhesive spray system for shoe outsoles and uppers. First, this paper describes how to extract the three-dimensional data of an outsole outline from a two-dimensional CAD drawing file. Second, it describes how to extract the three-dimensional data of an upper profiling line from the threedimensional scanning data that is opened in a three-dimensional CAD program. Third, it describes how to generate robot working paths based on the extracted data and the nozzle setting parameters for adhesive spray. Also, a series of experiments for adhesive spray is performed to verify the effectiveness of the presented methods. This study will do much for increasing productivity in shoe manufacturing as a core work of a robotic adhesive spray system.