Reflect Delay (2005)

Interactive video installation and custom software

  /    /    » watch video (large 117 mb)


Reflect Delay explores privacy issues raised by surveillance in an interactive video installation. The piece deals with issues of body, memory and gesture, as viewers are confronted with images of themselves in ways that disassociate them from their own body in a public setting, which has the trappings of private, domestic space.

Composing the piece suggested that interactive art can also be performative, requiring people to perform as a precondition of the piece. Reflect Delay confronts people with their own image in a public setting inspiring a number of different reactions. The piece inspired performances on the part of some and caused others to shy away completely. Because the installation allowed the rest of the viewers in the gallery to see the current reaction to the system without the superimposed image, we were able to observe people’s behavior with the system. One person sat for a long time in front of the piece without talking and moving while others used the system as an opportunity to perform in front of the group that gathered around the piece. One woman occupied the audience’s attention for a ten-minute unrehearsed performance with the system. Others interacted with each other, attempting to synchronize their movements so that their real and delayed images on the screen would dance together, or hide behind each other.

Reflect Delay makes use of a Brechtian interruption of chronology to pose questions about identity and memory. The method of the piece calls to mind aspects of the work of filmmaker Michel Gondry, who visited the Media Lab for several days shortly after the piece was completed. After spending an hour interacting with the work and discussing its meaning, he maintained that the piece successfully investigates questions of identity by using a visual doubling effect and went on to suggest that the piece might be even more effective were it to incorporate an audio component, which would allow participants to converse with themselves seven seconds in the past. We wondered together how the incorporation of audio in Reflect Delay might change speech, and whether the incorporation of audio in the piece would simply result in chaos or meaningful self reflection.

Regardless of the possible future addition of audio, Reflect Delay was considered a success in inspiring impromptu collaborative performances. Negotiating the conditions that would help people who were not otherwise performers become comfortable enough to act out in front of other people foreshadowed the preoccupation with barriers of entry that also govern the lasting relationships enabled by online social communities.

[ Top ]


Misty Dawn is a video installation developed with Philip DeCamp during the fall of 2005. Misty Dawn was installed in the Joan Jonas Performance Space at MIT in December 2005, at Art Interactive, a gallery in Cambridge, MA in February 2005, and at the MIT Media Lab, at a presentation for filmmaker Michel Gondry in April 2005.

[ Top ]


The custom software of Misty Dawn utilizes real-time image segmentation, superimposing video recorded seven seconds in the past onto video captured in real-time. The video software is an integral part of the installation; it consists of a projected image opposite a sofa and coffee table set, and is intended to evoke a domestic living room. A camera concealed in the coffee table points toward the sofa, recording the motions of gallery visitors who sit within the camera’s field of vision. The physical setting of the piece invokes a place where socializing occurs within a private home, so that would-be performers would feel less self-conscious as they interacted publicly with the piece. Visitors see an image of themselves projected on the screen in front of them in which the time-delayed foreground is layered on the current video to produce a doubling effect. The piece creates an uneasy spatial relationship in which viewers interact with versions of themselves seven seconds in the past.

[ Top ]

Technical specification

The program, written in C++, processes video input from a camera with an adaptive algorithm that can distinguish between the foreground and the background of the scene after approximately ten seconds of analysis with a steady camera shot. Following this analysis, the system continuously subtracts the background of the frames seven seconds in the past. The process produces a real-time video output in which the active foreground figure seven seconds from the past is superimposed on the present frame in its raw form. The real-time image segmentation generates fluid interaction between the actual person and her projected images.

[ Top ]