What is a real-time installation?

In the slew of sexy project write-ups and press releases about interactive and real-time installations, it can be hard to tell just what exactly a real-time installation is. I found a number of different definitions for the term, ranging from:

1. relating to a system in which input data is processed within milliseconds so that it is available virtually immediately as feedback, e.g., in a missile guidance or airline booking system. (Google)
2. of or relating to applications in which the computer must respond as rapidly as required by the user or necessitated by the process being controlled. (Dictionary.com)
3. the actual time during which something takes place (Merriam-Webster)

…So there’s even quite a bit of grey area in the definition of the term, let alone artistic uses of the term. I often break the term into sub-categories of real-time installations. The four areas I focus on are:

1. Generative installations
2. Interactive installations
3. On-demand installations
4. Self-managed installations

This helps clients, designers, and individuals outside the industry wrap their heads around the possibilities and potential uses of real-time installations.

Generative real-time installations

Generative real-time installations are often what wow people. These are the kinds of installations you see that are generating content, whether in the digital or analog world, within a handful of milliseconds. These kinds of installations are the kind that offer the possibilities of endless variations of content over the installations lifecycle, because the content isn’t a handful of static files. All the content is being created as you experience it.

The simplest forms of generative content can be thought of like music visualizers while the more complex content elements can be sensor-driven visuals (like Kinect points clouds) or an array of mechanical objects being controlled (like lasers or lighting fixtures) generatively.

I separate these from interactive real-time installations because there are many types of generative installations that don’t have any inputs or controls. Data visualizations or content driven by web API data would fall under this category. You may have even seen generative content used in commercial settings like stores that looked amazing but didn’t have any interactive element to them.

Interactive real-time installations

Interactive real-time installations are ones that feed off user interaction to change the content within a handful of milliseconds. This could be include gesture-driven menus, VR movies, or sensor-based installation controls. They still work within the same time frames as generative real-time installations but can have very different goals.

There are many real-time installations we’ve worked on that are real-time and interactive but don’t have generative content. They may have layers of pre-rendered video or complex sequences of events that are explored by users in real-time, but nothing generative at the end of day. These systems may have been considered generative in the past, but with the depth of what a generative system can do, I think it is more apt to separate them.

This doesn’t mean there aren’t a large number of real-time installations that combine both generative content and user interaction. The ones combining both are what make the waves and go viral. I do think it is important to present generative and interactive systems separately, since both can exist independently, and for a large number of clients, being able to imagine them separately gives them more possible use cases.

On-demand real-time installations

On-demand real-time installations are an interesting breed. The previous two types of installations had time frames within a handful of milliseconds to either generate content or process interactive elements. On-demand installations have a much larger window of “real time”. I would consider the time frame for on-demand installations to be in the 10s of minutes to even a day. They work by pairing offline processing/rendering and real-time installation systems/playback.

As far as computers have come in the last decade, there are still many processes which are cost-prohibitive to be done in a handful of milliseconds, such as working with big data or comprehensive web APIs. The on-demand elements of these installations can take tough tasks and process them over 10s of minutes and then delivers the output to the real-time side of the installation.

Imagine trying to crunch all of data available in Facebook’s graph API and presenting it in real-time to users in some form of visualization. It may be possible, but in the reality of installation work keeping costs down and timelines short can be the difference between getting a gig and not getting a gig. It may be preferable to have one computer ingesting data and creating visualizations as rendered content which can then be played back by another real-time systems that provide interactivity. To most users, this would still appear to be a real-time system as the data would still be updated regularly.

These are useful types of installations when considering large commercial roll outs, where a large number of low-cost devices can bring new content to the displays, while a smaller number of more powerful devices are rendering new content as requested. Live streaming of content from cloud render farms to Android-style playback devices has also become more and more feasible.

Self-managed real-time installations

The last category of real-time installations are self-managed real-time installations. This is a catch-all term I use for installations that have some real-time elements that drive their internal logic and have easy management abilities because they are “live”. I use the term “real-time” loosely because many of these kinds of systems end up being advanced video playback systems with some nice extra features instead of being awe-inspiring art pieces. There is definitely a living to be made making these kinds of systems even if they aren’t inspiring projects.

They may include elements such as extremely large video walls, auto-calibrating projection systems, or even systems that are easy to access through a CMS (think building management running monthly news on various screens in an office building lobby). These things may have no interactive elements, no generative content, or even any on-demand content, but they’re ability to be flexible in real-time allows them to excel in comparison to other kinds of systems.

TouchDesigner Installations

If you’re learning TouchDesigner and you’re curious about preparing TouchDesigner installations, we’ve got the answers for you here:

Wrap Up

The goal with this post isn’t to get a new definition of real-time installations because I think it’s time we move beyond that. We should start thinking of new and more accurate ways of describing installations and showing what systems are doing behind the scenes. I don’t think we’re doing any good by keeping all these various kinds of installations clumped under one umbrella term. I can’t count the number of clients to whom I’ve had to explain that it’s not an all-or-nothing situation. These are my four ways of breaking down real-time installations, what are yours?