In 3D graphics applications, effective camera placement and control is fundamental for the user to understand the virtual environment and be able to effectively accomplish the intended task. In most current 3D applications, users directly position the camera using a input device (which might require considerable skills and be error-prone) or simple automatic control methods (e.g. 3rd person cameras used in many videogames) are employed. In recent years, some researchers have come up with methods to automatically position the camera that aim at relieving the user from direct control, and are inspired by how human cinematographers approach the same problem: first, requirements on the image or shot to be obtained are stated (e.g., about size and framing of subjects), and then, using AI methods, a camera fulfilling those requirements is computed. Our research aims at: (i) finding efficient methods for automatic camera control that provide cinematographic-quality experiences and ease of use in complex virtual environments; (ii) experimenting those methods in applications for the design of 3D virtual environments and in serious games for training.