michael naimark

Tomorrow's Realities catalog,
Siggraph '91, Las Vegas


"VBK - A Moviemap of Karlsruhe"
Mediated Reality and the Consciousness of Place

Michael Naimark


I. Something Extraordinary Happens

It's last October. I'm in San Francisco sitting in a video editing room looking at the film transfer of the Karlsruhe footage, which I shot the previous month from the front of a tramway car. During the next few days I will edit it for a videodisc. I am watching hundreds of little snippets of straight and turn sequences whiz by at hypnotic speeds. For a few minutes it's kind of fun. Then it all starts to look alike. Dang, I can't believe I'm doing this again. It isn't quite like making a linear movie with an aesthetic of montage - the purpose is to organize the material to minimize disc search time and to make programming efficient. What a drag. But it's gotta get done.

Around the third day of nonstop logging and editing something extraordinary happens: I know where I am in the footage. Always. You can show me any of the forty-some thousand frames of the 108 kilometers shot in both directions and I can point to the corresponding place on a map, as long as I can "move" back and forth a bit. It actually seemed to happen all of a sudden, like I was astral-projected to see Karlsruhe from a God's-eye view.

I had formed a mental map. "Karlsruhe" became a singular thing, an almost living entity with which I could now relate. Once that happened, every frame had its place.

A colleague once told me he believed he could tell what section of Paris he was in purely by the quality of light.

I suspect this sense of wholeness, of consciousness of place, can be conveyed in a very fast and highly impressionistic way with such emerging interactive media. Not by simulating reality - a trap perpetuated by the believers in the objective and ultimately a losing battle. But by abstracting reality - creating experiences otherwise impossible in the real world. And that these experiences, when done artfully, will make you appreciate really being there even more.

II. Moviemap Basics

"My" definition of a moviemap is:

• user-controlled seamless navigation through a real or created place via optical disc;

• optical disc as lookup medium (no realtime image generation);

•user has realtime control of one-dimensional speed and direction;

• user has occasional control of two-dimensional choices but only at intersections (I call this "1.1D");

• there may or may not be additional non-seamless hypermedia information (ie., "destinations," tied together by the surrogate travel "routes").

III. Background

In late 1977, the first prototype laserdisc players were introduced to a small group of research institutions, including M.I.T.'s Architecture Machine Group. I recall the day it came (I was a grad student at the art center across the street, straddling between it and several labs). "ArcMac," at the time, was often viewed as a "computer graphics lab," but was more a vehicle toward understanding deeper processes, evolving out of Nicholas Negroponte's original credo that "computers should know their users." Past and current mega-projects at that time included "Graphical Conversation Theory," "Spatial Data Management Systems," and "Mapping By Yourself," so it was natural to investigate the videodisc's potential for making virtual environments.

The following spring, Peter Clay, an undergrad, shot some single-frame film footage travelling through the M.I.T. hallways with some help from Bob Mohl (a grad student who went on to write his PhD dissertation on moviemaps) and me. By the summer of 1978 we were ready to shoot something more. A real environment was selected - Aspen, Colorado (in part because of its distincitvely picturesque landscapes).

During 1978 and 1979 Aspen went through a quiet media "sweep." Under the direction of Andy Lippman and with additional help from wildlife cinematographer John Borden, cognitive psychologist Kristina Hooper, filmmaker Ricky Leacock, and others, streets were shot with 4 16mm stop-frame film cameras (pointing front, back, left, and right) triggered to fire every 10 feet by a fifth wheel on the back of our vehicle. The camera pod was stabilized by an expensive gyro platform. We also shot with a 360° fisheye-style lense. In addition to filming the routes, we shot stillframes of every facade in town (twice - both in summer and winter), stillframe "slideshows" of many interiors, short movies, and audio interviews. Rebecca Allen and Steve Gregory recorded binaural sound. Scott Fisher reshot historic photos from the same points of view.

Back at the lab, Steve Yellick digitally "de-warped" the fisheye footage. Also, the entire town was hand-digitized by Walter Bender into a crude 3D cartoon-like model. The "basic system" required at least two videodisc players both running into a switcher so that when one player was playing, the other was cueing, thus eliminating any blanking during searches.

The Aspen Moviemap was funded by the Cybernetics Technology Office of DARPA. Several other (mostly military) moviemaps were sponsored after, mostly (I was once told by one producer) "cheap and dirty" compared to Aspen.

In 1985 I directed production of a moviemap of a section of downtown Paris for the Paris Metro. With Bob Mohl, I shot using a custom 35mm film camera along sidewalks on a modified golfcart, triggering one frame every 2 meters. Rather than filming turns, we hired a mime to stand in the intersections and point. The system would then cut from the pointing mime to the direction she was pointing. The theory was to replace visual seamlessness with cinematic seamlessness (a la Eisensteinian montage theory). The Paris "VideoPlan" was on public exhibition at the Madeleine Metro stop for two years.

"Palenque" was filmed later in 1985 by the Bank Street College for the (then RCA) Sarnoff Labs as a prototype DVI application. Under the direction of Kathy Wilson, both Bob Mohl and John Borden helped shoot footage of walking trails. Palenque is a extensive multimedia package, including stillframes and text information about the site, as well as semi-realtime "dewarping" of fisheye images.

In 1987 I conceived and directed the "Golden Gate Interactive Videodisc," commissioned by Advanced Interaction, Inc., currently on display at the Exploratorium. A 10 by 10 mile grid at one mile intervals was carefully shot from a helicoptor with a special gyro-stabilized camera system, centered on the Golden Gate Bridge. The input device was a trackball, and with software designed by Ken Carson, created a feeling of realtime control, or "tight linkage," between what you do and what you see. This system also required two videodisc players and a switcher.

IV. Karlsruhe

The Karlsruhe Moviemap was commissioned by the Zentrum fur Kunst und Medientechnologie (ZKM), a state-funded arts and media lab under construction in the town of Karlsruhe, Germany. Karlsruhe has a well-known tramway system, with over 100 km of track snaking from the downtown pedestrian area out to neighboring villages at the edge of the Black Forest.

The entire tram system was shot in both directions from a tramcar outfitted with a 16mm film camera triggered by the tram's electronic odometer (at 2, 4 and 8 meters per frame depending on location). The tracks assure unrivaled stability and seamless match-cuts.

Using the tramway line as the basis for a moviemap of the town has its drawbacks. It doesn't go everywhere. For example, Karlsruhe has two large parks which are barely visible in the footage. Also, the presence of the rails, while perhaps adding a sense of visual continuity during travel, may distract the viewer from "looking around."

Yet the tramway routes are there for reasons of history, culture, politics, and geography.: not a bad basis for sampling the place.

The delivery system is controlled by a Mac II computer using Hypercard with software designed by Christoph Dorhmann. It consists of a large projected video image from a single Pioneer 8000 videodisc player (whose built-in frame buffer eliminates blanking during searches), a graphic map-and-cursor display, and a custom-built input consisting of a broomstick-size lever for controlling speed and direction (zero to mach three) and three footswitches (left, center, and right) for choosing which direction to go at each intersection.

The installation is intended to be transparent in its responsiveness (no significant lags) and culture-independent (e.g., no text). Each input device has an indicator light on it. When an input is active its light flashes until it is used. When it is being used is stays lit. When it is inactive the light goes out.

V. The Future: Shooting for Cyberspace

Immersion in ortho-stereoscopic imagery with unconstrained head motion and realtime manipulation is often considered the essence of "virtual reality" or "cyberspace." Today it is primarily restricted to the cartoon-land of computer generated images.

The future of moviemaps lie entirely in how they integrate into 3D computer models. Eventually, camera input will be used as the basis for such computer models (see M. Bove's PhD dissertation "Synthetic Movies Derived from Multi-Dimensional Image Sensors," MIT Media Lab 1989). Whatever was not shot will be interpolated, not an easy task, particularly when shooting in the field. Similarly, the issue of when to use a 3D realtime computer in the final delivery system and when to use a pre-stored version (or anything in between) will be a function of cost, state of the technology, and as always: content.