Desiging Parsifal, From First Meeting To Opening Night
Richard Wagner's Parsifal is considered by many serious opera fans to be the ultimate Wagner. Seattle Opera, perhaps the biggest champion of the composer's work in the US, chose this challenging tale to represent not only the completion of Wagner's canon, but to inaugurate the opening of the new Marion Oliver McCaw Hall. The new venue features great planning and execution both up and downstage of the proscenium, and the Opera was eager to create a production that met the scope and grandeur of their new home.
In June of 2002 we were approached by the opera to design the projected scenic backdrop for Parsifal. We were joining a creative team that included Robert Israel executing the production design, Michael Chybowski designing lighting, and direction by François Rochaix. Our first day-long meeting was a wild creative frolic. We all discussed the story, the music, the characters, and the mythos. Without restriction we began to feel out what this story meant to us, and how we might view its presentation. We left that day knowing that no matter what the outcome, the process would be interesting.
Because we were designing this first show, we also came to be involved with developing the projection plan and installation for the new venue. This happy correspondence allowed us to have an advanced view of what the production would have at its disposal. We worked with the venue to coordinate a series of presentations and product tests for systems that would be appropriate. Ultimately, Digital Projection Inc. (DPI) proved itself to be an enthusiastic and motivated partner in equipping the hall. Two Digital Projection Thunder 10k units were purchased for permanent installation in the venue. Chuck Collins, with DPI, worked closely with us and the city to ensure appropriate configuration and lens choices for the units.
This purchase decision was the cornerstone of the rest of the gear specified for Parsifal. Seattle's purchase of the Thunder units coincided with a significant investment in the same projectors by Scharff Weisberg in New York. With the knowledge that we would be able to assemble up to nine Thunders for the show, and with an adequate budget in place, we proceeded with confidence into production planning.
Our next meeting was with Robert Israel at the Seattle Opera scenic studio to view the model. Israel's inspired design provided a series of offset and angled planes. Levels of deck, arrays of walls, all backed by an enormous (38' × 80') projection drop provided a dramatic palette for the action of the opera. The story of Parsifal is Wagner's interpretation of the myth of the Holy Grail. The opera has always inspired ambitious and technically complex productions. Wieland Wagner's production in the 1951 Bayreuth festival featured an as-yet-unseen modern take on the opera, with abstract forms, and introduced, as the foremost design aspect, large-format film projections. In subsequent years, both Robert Wilson and Wolfgang Wagner went on to create their own technically and aesthetically adventurous productions of Parsifal. Our production would not lack for imagination and scope. How it would stand up to this prior lexicon would be seen in a year's time.
Our next meeting took place at our own Mode Studios in Seattle. We've configured the studio space to allow any of our editing, animation, or compositing stations to display output on a projector in our meeting area. With the creative team convened, we were able to view materials on a larger scale. We had converted the CAD drawings of Bob Israel's set into a three-dimensional previsualization using Softimage XSI. With views of the set pre-rendered with an alpha channel, we were able to put potential show images behind the scenery, simulate transitions, and get an idea of how the projected image would work with lighting looks and scenic configurations. We used a combination of Adobe After Effects, Photoshop, and Windows Media Encoder to generate the previews and to alter projection images in real time during the meetings. This method of pre-visualization and visual brainstorming proved to be successful and liberating. With such a clear picture of what was possible, we were free to explore different treatments.
Ours soon became a world of complex, interrelated numbers. Dimensions, aspect ratio, blend percentage, pixel resolution, and processing ability were now the order of the day.
We had several dimensions that were known from the start. Accounting for the length of the projector, we would have 17' from projection screen to projector lens. In order to fill a drop to the desired dimension we were clearly going to have to blend projectors. Consultation was required, so we repaired to New York to meet with Peter Scharff and Lars Petersen from Scharff Weisberg. Both expressed some concern at the idea of blending in two axes, both horizontal and vertical (see figure below). When blending, the area where the two projectors overlap is receiving twice the light output, and thus can be difficult to obscure. Peter and Lars felt strongly that adding an additional tier of projectors would create several areas where four projectors would be overlapped; and that these areas could prove particularly problematic. The resulting blend could have a difficult time if the art were too dark, too light, or too consistent. Peter suggested that we consider one tier of projectors turned on their sides. With the larger dimension of a 5:4 ratio projector turned vertical, we would achieve almost 30' of vertical image. We were persistent in needing to get 38' of vertical dimension though, and we were quite ready to throw caution to the wind to get it. Our creative meetings had shown us that perhaps the art would be forgiving. Thus to achieve the image size, we eventually arrived at eight projectors (two tiers of four) with 20% overlaps. This made for a native image size of 4160×1792. Although much of the opera would utilize still images, there were plans for several large-scale image effects that would be full motion. We would be asking a great deal of the Dataton Watchout system in processing frames this large at full speed. We left New York determined to do test renders to discover how large we could go with full motion, and what CODECs would be most efficient.
This fixing of what the gear would be, and what the pixel size of the resulting blended image was, allowed us to begin creating full-resolution versions of the art suitable for the show. We had come away from the last creative meeting with a consensus on several directions to pursue. We now began the process of creating the media required for the show.
Bob Israel had shown several concept images in our initial meetings. The first was a picture of a snowy plain with a cluster of trees at its center. We quickly recognized this location, a beautiful spot in the mountains near Seattle. The snowy plain image would be used in Act I, and in Act III, at which time it would be required to undergo springtime rejuvenation. With the location nearby, we made a trip out and were able to find the precise point from where Bob had taken the picture. This location was in the middle of a dry lakebed, accessed from the side of the freeway. We drove a fluorescent orange stake into the ground at the location of the tripod and shot a series of panoramic shots of the location with a four-mega-pixel digital camera, as well as shooting coverage with a DV camera. With our location neatly marked, we were able to revisit the exact location several more times as spring took hold, gathering a series of shots in a wide variety of states, from snow covered to full bloom. By morphing through this series of images, and careful compositing we planned to have the location actually thaw and bloom over the course of three minutes in the third act. This would correspond with the “Good Friday Spell” that defines the turning point of the act.
The second image was a picture of Mt. Whitney that Bob had shot from his car. The image was also to be used in Act I and Act III to designate the location of Tintagel, the grail knight's castle. We couldn't obtain the negative of Bob's original image. The picture had been taken in high summer, and it was now the winter before the show. It would be impossible to recreate this image before the opera actually had to take place. We turned to another local resource here in Seattle, Getty Images. We were able to license a stunning high-resolution picture of the same location that suited our needs perfectly. With this stock photo as a starting point, Colleen took the lead in developing a series of painterly and somewhat ominous versions of the location. Digital lighting effects were added to the mountainous peaks, with color correction and grading accomplished using Photoshop and After Effects.
The second act took us out of these locations, and into the realm of the evil wizard protagonist, Klingsor. Our production of Parsifal had cast Klingsor as a mad scientist, a man who creates his own world through code and scientific method. We thought it would be interesting to give the Act II projections a certain sense of “computerization” to reflect Klingsor's manufactured atmosphere. We pushed colors to be wildly saturated, and used certain deliberate pixelizations to represent the imperfections of Klingsor's code. Ultimately when he his defeated, Klingsor's world would devolve to static. This ending would come with an epic five-story plunge of his onstage tower.
These Act II effects necessarily included a great deal of motion. We created the images and sequences in After Effects. Since it would be impossible to know what resolution would adequately play back in given sequences until we had the full Watchout system, we rendered multiple versions of all. These multiple renders included uncompressed QuickTime at full resolution and uncompressed at 1/2 resolution. Photo JPEG Quicktime files at full and 1/2 resolution were also prepared with a moderate amount of compression.
Ultimately we discovered that playback performance depended to a large degree on the nature of the content. Some of the kinetic footage mainly consisted of long transitional blurrings, or color shifts, and these proved to have viable playback at full resolution, uncompressed. Imagery that either had fast moving elements (particularly linear horizontally moving elements) required that we use the Photo JPEG flavor of the Quicktime CODEC at 1/2 resolution (2080×856). When scaled by the Watchout system, the footage proved to still be quite crisp, and the playback was flawless. Oddly enough, the most difficult clip to playback proved to be the static at the end of Klingsor's demise. Static is highly kinetic in a video sense, and the bitrate for the video proved so large that we ultimately used the Watchout to scale a 1/3 resolution render to achieve smooth playback. Fortunately, because it was static, the resultant image degradation looked natural.
We continued this development path for the intervening months, with the astute guidance of Robert Schaub, the Seattle Opera's director of technical production. He was diligent in bringing the design team together on a regular basis, and in setting up a pipeline for production information and planning. Schaub's gently firm hand kept the production moving towards a successful conclusion. These meetings and communications brought all of us into a tight consensus of what we aimed to achieve in each scene, and what palette of looks we would hone and bring to the stage for production. We were very ready for tech when load-in commenced in the new theatre on July 8th.
Because of our extremely compact throw dimension, we were compelled to use .8 lenses. This introduced a complication into the blending. Due to the inherent optical aberration that occurs in these extremely curvaceous short throw lenses (an aberration that results in a “pulling” of the image at the corners), we would have to place the projectors as precisely centered as they could be behind their display areas. It was critical that we not utilize optical lens shift as much as possible, as it would exacerbate the already present aberration. Tim Buck, Seattle Opera head carpenter, devised a rigging scenario that would place the projectors at these exact locations. The trusses were fitted with custom hardware that allowed for leveling, physical focus, projector lockdown and the mounting of accessories. This truss array came together with great speed, the projectors were quickly mounted, and the array rose. We had specified color scrollers with gradiated strings of Rosco neutral density gel to be mounted in front of the projectors. This discreet system would allow us to ‘douse’ any projector, regardless of whether or not we may have electronic control of it. Finally, we had also specified a system of 14 Martin Mac 2000 Performance units with custom high-resolution gobos and effects wheels to layer with projection, and to serve as a backup system in the case of a projection system failure. Bob Schaub had been clear that the Seattle Opera required such contingencies to be addressed proactively, and we cued these alternative scenarios in the same tech sessions as the primary media.
Bill Gillette, Digital Projection's top projectionist, accompanied by Tom Whipple of Scharff Weisberg, led the local projection crew through the initial alignment of the projection array. The individual units were aligned using electronic image positioning, keystone correction parameters, and signal amplitude parameters. Utilizing light meters, the array was balanced for luminance and color. Once alignment and balancing was achieved (a process that took about five hours initially), Lars from Scharff Weisberg worked to develop an optimum blend ratio for initial cueing on the Watchout system.
The final blend curve would be adjusted prior to each show as it turned out, and didn't get close to its meaningful resting place until we began to see the blend in context with the light cueing. By adjusting this blend curve parameter we were affecting the luminance gradient that cross blended the edges of the separate projections together. Adjusted too far in one direction and you see bright luminous “bars.” Too far the other way and you get a grid of shadows. As the projectors burned in over the run, the curve was adjusted to accommodate the gradual change in lamp color temperature. The blend adjustment works differently with different art, bright consistent images tended to look better with a different blend ratio than the dark, brooding, or colorful images. One thing that immediately made our wish list for future versions of Watchout was a key-frameable adjustment of this parameter that would allow it to be changed over time within the show to optimize for different sorts of art. At this point the system is forced to endure a processing period while the new blend adjustment ‘takes’ during which your screen blares forth with the Watchout logo. It became something of a ritual for the director, François, to duck and cry “Watchout !” when we would update the timeline information. It's probably safe to conjecture that as host platforms for Watchout get faster, that these interruptions will become more rare and certainly briefer.
With the system adjusted we began to load media onto the Watchout master node. Once media that is contained on the master is placed on the timeline, it is distributed to the individual processing nodes (one per projector). These nodes then assume the bull work of displaying the appropriate part of the image, as well as calculating and rendering the blend. Once this distribution of media is achieved, the timeline plays back media in real-time! We ultimately discovered that we could reliably run four image tracks simultaneously, as long as no more than two of them had moving imagery. With still cross-fading we used up to six active tracks. This real-time flexibility made changing timing, and image attributes as responsive as cueing lighting in many senses. We sat at the production table with the master node display, keyboard, and mouse, as well as our own custom created laptops. These laptops are equipped with Avid editing systems, Adobe Photoshop, and After Effects, as well as a slew of other tools. They are quite nimble and speedy; and networked with the Watchout, we were able to make significant aesthetic changes to art, download to the Watchout, and view it onstage in minutes. This speed and flexibility was very, very, very satisfying.
As tech proceeded we worked closely with lighting designer Michael Chybowski to match the tonality and temperature of the lighting onstage. Bob Israel asked for the background images to gradiate into the stage look as much as possible in most cases. The aforementioned system flexibility allowed us to do this on the fly. We developed a method of mixing warm, cool, blurred, and luminance varied versions of the art in the tracks to make interactive adjustments to color and focus (see figure above).
Tech proceeded smoothly, almost too smoothly, and we experienced two giddy days where we were able to complete Acts I and II. On the third day we began Act III, and the inevitable bumps in the road commenced. We had collectively decided that the physical melting and blooming of the lakebed in Act III was too realistic. The scene needed to be less naturalistic and a bit more mystical. We flailed about looking for a solution, and left the theatre that day unsatisfied. The next day was an official day off; nevertheless, we convened with Bob Israel to brainstorm. A strong pot of coffee and some delicious scones eventually pointed our brains towards a creative and elegant solution. Happy with our results, we were able to set the studio machines to render the new sequence, and get some rest. Within days our happy tech process was complete, and the show had opened to wonderful critical notice.
During the course of the run the Thunder 10k projectors remained marvelously consistent and problem-free. Any small issues that did crop up were solvable in a timeframe and effort that would be normal for any pre-show lighting checkout. Each day, the newly minted local projection crew (the very talented Chuck Whitmore and David Misner) would realign the array, accounting for movement of the screen due to weather, and to the inevitable minute thermal shifting that happened during the show's run. It eventually proved to be viable and advisable to touch up this alignment between acts, as the presence of epic scale scenery in different positions caused the vast projection drop to billow slightly in different directions in each act.
Upon completing the process, the creative team had just scratching the surface. The unbelievable sense of flexibility, coupled with the power of the medium had made a huge impact. It is easy to see that as the computing power of the distribution and manipulation gear grows, and as the interfaces for these things become more intuitive, the projection toolset will become pervasive. The economy and reliability have already made such strides forward that we felt confident and able to do this ambitious project. Certainly without the generous help of Digital Projection and Scharff Weisberg we would have been up against a very difficult task. But with the production behind us, we are now determined to try some larger and more ambitious blending arrays. As designers we will continue to be most satisfied when we've made our vendors just a little bit scared of what we want to attempt!