In this work, we propose the use of a Natural User Interface (NUI) through body gestures using the open source library OpenPose, looking for a more dynamic and intuitive way to control a drone. For the implementation, we use the Robotic Operative System (ROS) to control and manage the different components of the project. Wrapped inside ROS, OpenPose (OP) processes the video obtained in real-time by a commercial drone, allowing to obtain the user's pose. Finally, the keypoints from OpenPose are obtained and translated, using geometric constraints, to specify high-level commands to the drone. Real-time experiments validate the full strategy.