The director of an event held annually in Las Vegas on New Years Eve asked for help mapping video onto some costumes that would be onstage for about 30 minutes during the opening walk-in numbers. After assessing projection angles and maximum effectiveness for stage lighting, I convinced him to position a performer on the top center platform of the stage so that I could project in 360° onto the performer’s dress, something visually exciting and unique for the guests.
The Gear
After consulting with Ruediger Haeming from Barco/High End Systems, I decided to use four of their DLHD digital light fixtures based on their collage-building and circular warping capabilities.
Ruediger and High End also provided support during the planning stages and also onsite, easing any concerns I may have had with networking setup and new features in the Axon HD media server, Hog4 Full Boar console and CMA (Content Management Application) that I relied upon for this project.
Design Considerations
I had to make sure that the dress was covered from the shoulders to the ground with as much intensity as possible. This meant finding the throw distance that would give us the image size we needed with as little waste as possible. And since I was also the lighting designer for the event, I was able to hang the DLHDs in the best possible locations to insure maximum visibility of the projections, intensity and image size.
Challenges
The fabric of the dress was uneven, and the shape of the dress was flared. This meant that some areas of the dress would likely be slightly out of focus unless we could minimize the folds in the skirting. To the eye, from about 25 feet away, however, any issue with focus was imperceptible, so we settled for a sharp focus on as much of the dress as possible.
Visibility and sight lines: Since the performer was positioned on the top center platform of a stage in the round, sightlines were in our favor. The dress could be seen from all sides at every angle.
Movement of the performer: The performer was told to not to move her feet, and to limit her side-to-side movement somewhat. Otherwise, she was able to move her arms and head, and was even able to rotate in place slowly without any real issue.
Solutions:
We created a content management network between the four DLHDs and an Axon HD. This allowed us to use the Axon HD as the main CMA interface for content management and create the warps on each DLHD directly.
To identify and correctly assign each piece of the collage to a DLHD, I created a test image with four vertical sections, numbered 1-4, and each section in a different color. Then we projected this image onto the dress, quickly and easily assigning the correct portion of the collage to the appropriate DLHD from the Hog4.
To correct the keystoning issues that result from projecting 4:3 rectangles at a 45° angle onto a vertical object, we used a test image with straight horizontal lines (in this case, the color bars rotated 90° worked perfectly). We then worked our way around the dress, one DLHD at a time, making adjustments to the points in the warp mesh of each of the four DLHDs until a horizontal line continuously wrapped around the dress and blended correctly into the next DLHD.
To see the projection firsthand and make needed adjustments on the spot, we positioned the Axon directly on the stage on a rolling case and moved around the dress, viewing it from all four sides. It was so much easier to get the vertical and horizontal alignment and spacing between pixels correct this way; because I could see it firsthand, there was no guesswork needed.
Overall, creating the warp and using the collage feature of the DLHD was very easy and straightforward. As we worked through creating the warps in the CMA on the Axon HD, though, it’s important to note that we paid close attention to the spacing between points in the mesh as we adjusted each one. Any slight adjustment to one point to level out a horizontal or vertical line usually meant adjusting the adjacent points as well. So, when creating a mesh for warping, it’s best to keep the spacing between vertices consistent.
Since the mesh is the map for how the video will be projected onto the object, the vertices in the mesh should not be distorted or contorted in drastic proportions. And another point to make here about the vertices in the mesh is that they cannot overlap! If points overlap, it will cause undesirable results with the mapping of the video (reversed image, duplicated image, or no image at all).
After taking these steps, we made the dress come to life. The images projected from the DLHDs were bright and they illuminated the dress with a continual sequence of
images — flowers, butterflies, crumbling blocks, a carousel, colorful sparkles and geometric patterns — as the guests entered the party. The animated costume also tied in with the scenic elements of the room.
I’d like to offer my thanks to programmer Matt Artigues for his programming assistance with the DLHDs and to Ruediger Haeming from Barco/High End Systems for all of his support.
