In this thesis we present software framework for real-time augmented realty applications. The framework is based on face recognition and capable of tracking multiple faces at once. It detects facial features and estimates position of the face relative to camera. We implemented a practical example that uses framework in an application used for marketing. Application requires a screen and a camera, so that the users can see themselves on the screen. Application draws some predefined object on the user's face, for example glasses or mustache, and also a comicbook-like cloud near the face with chosen text.