We use the front facing camera in a smart phone to capture gesture input. Thumb gestures performed above the camera are recognized and used to invoke commands. In contrast to other input modalities the camera requires no device movements and no valuable screen space is used. To be viable, this type of interaction requires gestures which are comfortable and memorable for the user and real-time accurate recognition of those gestures. Given the performance constraints of phones and their cameras we needed to determine whether accurate and reliable recognition is possible and identify types of gestures that are recognizable and user appropriate. As a proof of concept, we conducted a user study testing three gestures for performance and user satisfaction. The results demonstrate that the 3D gestural input is successful and we provide detailed insights into successful recognition strategies for this novel interaction modality.