This paper develops techniques using which humans can be visually recognized. While face recognition would be one approach to this problem, we believe that it may not be always possible to see a person's face. Our technique is complementary to face recognition, and exploits the intuition that human motion patterns and clothing colors can together encode several bits of information. Treating this information as a "temporary fingerprint", it may be feasible to recognize an individual with reasonable consistency, while allowing her to turn off the fingerprint at will. One application of visual fingerprints relates to augmented reality, in which an individual looks at other people through her camera-enabled glass (e.g., Google Glass) and views information about them. Another application is in privacy-preserving pictures - Alice should be able to broadcast her "temporary fingerprint" to all cameras in the vicinity along with a privacy preference, saying "remove me". If a stranger's video happens to include Alice, the device can recognize her fingerprint in the video and erase her completely. This paper develops the core visual fingerprinting engine - InSight - on the platform of Android smartphones and a backend server running MATLAB and OpenCV. Results from real world experiments show that 12 individuals can be discriminated with 90% accuracy using 6 seconds of video/motion observations. Video based emulation confirms scalability up to 40 users.