Sniff is created with Blender3d, and Unity3d Game Engine which renders the dog in real time and allows to dynamically change his behavior based on the video tracking data. The sidewalk is illuminated with infrared lights. An infrared-sensitive camera is used to monitor the sidewalk in front of the display windows. We track the position of each viewer and implement simple gesture recognition, so that fast, big actions are interpreted as threatening and slow actions directed at the dog (for example hand extended in his direction) are interpreted as friendly. The dog keeps track of the attitude of the viewer and forms a relationship with them over time based on the history of interaction. Sniff, simulating the visceral satisfaction of reality's responsiveness and dynamism, is an exploration of engagement, and of relationships created by one's presence in an environment. The experience is very familiar yet strange, leading us to re-examine notions we take for granted. Dog's behavior externalizes the process of assessment, evaluation and testing we perform every time anything new enters in the scope of our experience. Sniff has us unwittingly enter into an exchange simply by following the basic instinct of stopping and looking at something that is paying attention to us. A tension is produced by a mixture of fears and expectations, curiosity and interest. Sniff is an interactive projection in a storefront window. As the viewer walks by the projection, his movements and gestures are tracked by a computer vision system. A CG dog follows the viewer, dynamically responds to his gestures and changes his behavior based on the state of engagement with the viewer.