Virtuality is an innovative process seen in recent years. This process allows transforming scenarios to a virtual environment for simulations, practices, or tests so that all results can be studied. In turn, these environments can be supported by new tools and methodologies that increase the analysis capabilities, such as motion capture to generate animations, tracking of robotic equipment and the study of human movement, which benefits the development of systems with these objectives. A prototype of a nonintrusive virtual trainer capable of capturing motion and determining the correct execution of exercise routines is presented, using a virtual environment developed in the Unity video game engine. The system employs a SHDR webcam for real-time capture of the movement performed by the user, which is processed to track the pose and joints using Machine Learning through the MediaPipe library. This article explains the construction of the prototype and presents the results of the project.