The use of computers and artificial intelligence (AI) has become firmly established in everyday human life. Computer-assisted surgery is already an integral part of many medical applications. AI is being used successfully to help with medical decision-making and to support various medical processes and interactions. To provide reliable assessments, AI systems must be trained on realistic and representative datasets. In sensitive domains such as medicine, there is often an insufficient amount of high-quality data available for effective AI training. Therefore, in this study, we model and prepare synthetic training data of a human knee for further use, aiming to eliminate the need for physically installed markers. We investigate the application of synthetic AI training data in the field of surgery, using total knee arthroplasty as a representative example.