Notes on the Evolution of the Live Video Delay System
Alex Carpenter (2010)


The Live Video Delay System essentially realizes a visual analogue to an improvisational approach I've been using in solo audio performance scenarios since 2003 (e.g. Deep Golden Flourish, Excavation Patterns). These scenarios involve one or several long audio loops (5-23 seconds), which begin as near-blank “canvases” and are built on in real time via a delayed playback/response cycle.

This process is different to perhaps more standard layered looping processes in that it does not begin with a structural foundation in place, or in mind. Rather, the structure evolves from – and is determined by – the layering process itself.

I like to think of this performance activity as a kind of “tending.” It is not the willing-into-existence of an already known thing, but a careful and reverent listening and responding, which ensures the thing grows from, and stands on, its own foundations.

What results is a growing sonic “organism” that is neither purely a product of the performer, nor completely reducible to the processing mechanism. The performer contributes new material, but he/she is always guided by what is played back, and what is played back is always the sum of his/her previous recorded “guidances.” In another sense, you could say the organism's path of development is shaped by its own previous incarnations and memories, and the impression they have upon the observing performer.

Taking this analogy further, perhaps there is a kind of dangerous seduction at play between the organism and the performer. The organism needs to seduce the performer in order to grow and flourish, but peril is always present in that the performer may at any time obliterate the organism, either by not responding sensitively or by exerting his/her power to reduce it to his/her own will.

The idea of utilizing this way of working visually has always appealed to me, but for a long time the conceptualization of a problem-free visual system eluded me. I knew I could probably realize something similar by editing video in layers, but this would preclude the possibility of real-time reaction, which was a vital part of the process. I also suspected there may have been a way of realizing the process through graphics software, which I wasn't exactly closed to (and still aren't), it just seemed to remain a last resort in my mind.

Mainly, I wanted the system to be highly responsive and able to capture nuance. I'd learned from my audio work that subtle differences in pitch, articulation, intensity, etc. often played a vital role in setting the course of structural development, as if in these nuances the organism-to-be suggested itself. So it was important the visual system was similarly responsive. Digital work flow is indirect by nature. It is not that nuance cannot be achieved, but that the route leading there is often long and convoluted, with many stages. In my case, this posed too much of a threat upon the required immediacy and spontaneity of response.

My ideal was really some form of “drawing” that could be captured, looped and built upon. I eventually realized the rudiments of such a system could be achieved with two camera-projector chains; one dealing with the feedback loop, the other dealing with the addition of new material. The feedback chain's camera would simply need to see the screen, then output imagery back to the projection surface after a certain amount of time (via a delay line). The second camera would need to see the artist's canvas and output imagery (with no delay) to the same projection surface, thus infiltrating the loop. But several problems remained. Aside from the artist's hand/pen being in shot, there were issues of visual clarity to address.

Layering projected video is not like layering sound. A sound in an electronic signal chain stands out against a backdrop of (virtual) silence. But a projected video subject (e.g. a dot) does not stand out against a backdrop of “no light.” Even a black background is made up of light. Thus, each additional layer would undoubtedly result in a greater build up of artifacts, noise, and in the case of brighter backgrounds, colour saturation. In other words, contrast and subject-definition could very quickly be compromised.

The task then essentially became figuring out how to impede the build up of artifacts as much as possible. Inverting the white background / black subject paradigm seemed to be a step in the right direction. But the real breakthrough came when I realized an even sharper isolation of the subject could be achieved if it was drawn in a more vivid colour and then captured through a corresponding colour lens filter. Colour lens filters pass certain colours while blocking others. This means a heavy red filter would pass a red subject with great clarity along with red noise components only, significantly reducing artifacts overall. Using red filters at both stages of capture (and possibly projection also), as well as increasing camera exposure, seemed to promise a more than workable system.

The idea of using a focused red laser as the image source seemed to not only reinforce this goal of artifact-reduction, but also resolve the hand-in-shot issue, and eliminate the need to deal with line-trail (i.e. having to clear the canvas each time new material was added).


How it works: Camera 2 sees the performer’s laser activity on the black surface and when Fader 2 is opened the image is projected from Projector 2 onto the projection surface. Camera 1, which is trained on the projection surface, sees the image and sends it to the DVR. The DVR delays the signal by a given period of time, then sends it to Projector 1. Projector 1 projects the image onto exactly the same projection surface, the image is seen by Camera 1 and the whole delay cycle begins again.

In effect, what is set up by Chain 1 is a “loop,” even though technically there is never any true repetition of data. Camera 1 sees at first a performance, then a playback, then a playback of the playback, then a playback of the playback of the playback, etc. If the performer was to stop contributing new material (via Chain 2), this “loop” would essentially be repeated continuously with no change other than generational loss.

The monitor and Fader 1 are both necessary supplements to the DVR. Originally I imagined the signal delay would be provided by a standard video delay line, but when I learned such units cost upwards of $1000, Jim Lyons at Allen Avionics in Mineola, New York, suggested I try a DVR. He, and others at Allen, discovered it was possible to set their stock model DVR to play the same file it was recording from any point along its recorded timeline, making the machine a very flexible and cost-effective delay line. However, setting up the playback process requires some initial access to an on-screen menu. Since having this menu appear on the main screen would be both distracting and detrimental to the loop, it was essential for Chain 1 to include a monitor and a fader. The playback can thus be setup, and the menu display switched off, before the stream is passed to the projector. The fader also acts as a master control, and allows the piece to be faded out at its conclusion.

The first test session of this system took place on September 21-22, 2009, at the MELA Foundation in New York City, where I regularly volunteer. I had acquired all necessary components other than the two projectors, which MELA very generously loaned to me. The results of this session were astounding, although not entirely as predicted.

Decay & Depth
Initially, I was disappointed to learn the precise wavelength of the red laser’s output was being distorted somewhere in the circuitry of the camera, resulting in too significant a loss of brightness through the projector’s filter. Essentially, the image, by the time it reached the projector, was no longer a pure red, and thus could not pass through the projector’s filter unaffected. But I soon learned such filtering was basically unnecessary, since, surprisingly, the camera filters provided more than enough isolation of the subject.

Still, colour was being very slightly compromised somewhere in the chain, and as a result the image was losing a small shade of brightness each time it passed through the camera’s filter. It struck me, though, that this “decay” was not necessarily a flaw. Although something was lost in terms of the longevity of the loop, something was gained in terms of visual depth. Instead of all layers remaining at equal brightness, older layers seemed to retreat further back and occupy a different plane to newer ones. An extra dimension of depth presented itself.

I found the extremity of this decay would change depending on the projectors’ brightness and contrast settings, camera exposure levels, etc., so the longevity of the loop could in effect be “adjusted.” I have no doubt the limits of this function could be expanded further still given better quality cameras (and thus increased accuracy of colour-rendering), the ability to perform manual white balancing (not possible on the cameras I was using), as well as different grades of lens filter.

Ideally, the system would allow me the flexibility to eliminate decay altogether, for the benefit of working on large- as well as shorter-scale structures. With my current equipment this would come at too great a cost to image clarity, but I feel confident I can gradually approach this ideal with better system components.

Generational Loss
Another factor which had a greater-than-predicted effect on the overall process was generational loss – that is to say, the loss of information caused by the continuous record-playback cycle. To better explain what I mean by this, it helps to reiterate the system does not technically involve “looping” at all. What appears to be a repeated sequence is in fact an entirely new recording of the previous recording – projected onto a screen, then seen and captured by a camera all over again. Each time the image passes through this cycle it loses an extra degree of detail. But, as a result, new lines, shapes and patterns are actually emphasized.

Since these new emphases are seen at the same time the material is being built upon, it is fair to say generational loss in this system is not merely an interesting effect, but can be a key contributing structural factor, by inspiring certain reactions in the performer which might otherwise never come to mind.

In this example, the detail of the dots is compromised to such a degree that they begin to merge together creating a solid mass of light, from which some dots occasionally break away and form smaller cells. Most new dots either follow and emphasize this breaking-away gesture, or hover in the center of the mass maintaining its brightness. Many also float around the edge of the mass changing its profile. These live actions engage with the effects of generational loss and become the content of a growing compositional structure.

I was delighted to find the system was capable of capturing even more nuance than I had imagined. This was largely due to the enormous flexibility of Camera 2, which, because of its heavy red filter, did not actually require a fixed background view, or even one that was consistent in colour (since anything moderately dark registered simply as black). This meant a number of features were suddenly very elastic, such the angle of the black surface in relation to the camera lens (affecting the shape and perspective of the dot), as well as the camera’s focus and zoom (affecting the detail and size of the dot).

Another type of nuance whose importance was confirmed in this session, and particularly in the above piece, was the “mistake.” Since everything captured by Camera 2 automatically becomes part of the loop for at least 10 generations, any given performed action can have a significant effect on future actions and thus the direction of an entire piece. I couldn’t count the number of times the smallest “accidental” dot sent a piece careering in a completely unexpected direction. In this particular case, it was merely a momentary blocking of the laser (with my finger) which led to the epic dance of flickering, multi-sized dots you see.

Obviously, any “mistake” like this could easily be ignored in its early stages and thus prevented from affecting the existing material in any significant way (it would eventually fade out). But I think this would be missing a unique opportunity the system presents.

As the drawer, I am tending to the growth of something, perhaps in a small way like a parent tends to the growth of a child. If a “mistake” happens to occur, it unavoidably becomes part of that something, in the same way a child might reveal certain interests its parent never foresaw. No loving parent, in my opinion, ignores or suppresses the interests expressed by their child, even if they happen to be incompatible with their own. They acknowledge and encourage them – give them attention – and as a result the child is allowed to flourish, and be in a way that is consistent with its own self. In the same way, mistakes reveal potential for the organism; they present new dimensions and suggest new ways it could be. Since the organism’s growth is guided almost exclusively by unplanned suggestions, it could even be said that the “mistake” is the purest expression of its very life-force. It seems for the organism to really flourish and develop according to its own self, the drawer must acknowledge and explore such “accidental” occurrences, not ignore them in pursuit of an already planned outcome.

But even this is not necessarily a matter for the drawer to simply decide. There must always first be a seduction. The organism must, in effect, make the drawer fall in love with it, if it has any hope of securing a self-consistent path of development. (At least, this is my romantic way of explaining the situation; obviously I realize the “organism” is not sentient.) Essentially, what I mean is there must be a point at which the drawer’s interest/curiosity in the growing organism reaches such critical mass that his/her own personal likes and inclinations are temporarily eclipsed by the needs of the growth process itself. The more the organism is allowed to show itself, the less the drawer has to think rationally about what to do, what to contribute. Only once this level of communication and bond is established, and the process becomes very fluid and intuitive, can the organism’s development begin in earnest.

For providing assistance and encouragement in the developmental stages of this project, I would like to thank Marcia Woodfield, Dana Jabeck, Sam Roden, Nick Hartanto, Jim Lyons, Allen Avionics, James Ross, Rob Ward, La Monte Young, Marian Zazeela and Jung Hee Choi.

© Alex Carpenter, 2010

Please visit the front page for details of the Live Video Delay System's latest developments.