Manual and camera-based game control is described. In one embodiment, a game system receives an image stream depicting a player of a game and also receives manual input from a hand operated controller used by the player. In an example, the manual input may be used to control an avatar and the image stream is used to recognize gestures of the player and control background objects in the game using the recognized gestures. For example, by using gestures such as head rotations the player's hands are free to continue operating a hand-held controller and the player has increased control and an immersive game play experience. In various embodiments, methods for detecting head gestures from image streams in real time are described such as tracking velocity of player's heads and motion through quadrants of an image stream.