Krakatoa Creates CG Visualizations for Short Film Rebirth
Film director and cinematographer Patryk Kizny - along with his talented team at LookyCreative - put together the 2010 short film The Chapel using motion controlled HDR time-lapse to achieve an interesting, hyper-real aesthetic. Enthusiastically received when released online, the three-minute piece pays tribute to a beautifully decaying church in a small Polish village built in the late 1700s. Though widely lauded, The Chapel felt incomplete to Kizny, so in fall of 2011, he began production on Rebirth to refine and add dimension to his initial story.
Exploring the same church, Rebirth comprises three separate scenes created using different visual techniques. Contemplative, philosophical narration and a custom orchestral soundtrack composed by Kizny’s collaborator, Mateusz Zdziebko, help guide the flow and overall aspirational tone of the film, which runs approximately 12 minutes. The first scene features a point cloud representation of the chapel with various pieces and cross-sections of the building appearing, changing and shifting to the music. Based on LIDAR scans taken of the chapel for this project, Kizny generated the point clouds with Thinkbox Software’s volumetric particle renderer, Krakatoa, in Autodesk 3ds Max.
“About a year after I shot The Chapel, I returned to the location and happened to get involved in heritage preservation efforts,” Kizny explained. “At the time, laser scanning was used for things like archiving, set modeling and support for integrating VFX in post production, but I hadn’t seen any films visualizing point clouds themselves, so that’s what I decided to do.”
EKG Baukultur an Austrian/German company that specializes in digital heritage documentation and laser scanning, scanned the entire building in about a day from 25 different scanning positions. The collected data was then registered and processed – creating a dataset of about 500 million points. Roughly half of the collected data was used to create the visualizations.
Data processing was done in multiple stages using various software packages. Initially, the EKG Baukultur team registered the separate scans together in a common coordinates space using FARO Scene software. Using .PTS format, the data was then re-imported into Alice Labs Studio Clouds (acquired by Autodesk in 2011) for clean-up. Kizny manually removed any tripods with cameras, people, checkerboards and balls that had been used to reference scans. Then, the data was processed in Geomagic Studio to reduce noise, fill holes and uniformly downsample selected areas of the dataset. Later, the data was exported back to the .PTS ASCII format with the help of MeshLab and processed using custom Python scripting so that it could be ingested using the Krakatoa importer. Lacking a visual effects background, Kizny initially tested a number of tools to find the best way to visualize point cloud data in a cinematic way with varying and largely disappointing results. Six months of extensive R&D led Kizny to Krakatoa, a tool that was astonishingly fast and a fraction of the price of similar software specifically designed for CAD/CAM applications.
“I had a very basic understanding of 3ds Max, and the Krakatoa environment was new to me. Once I began to figure out Krakatoa, it all clicked and the software proved amazing throughout each step of the process,” he said.
Even with mixing the depth of field and motion blur functions in Krakatoa, Kizny was able to keep his render time to roughly five to ten minutes per frame, even while rendering 200 million points in 2K, by using smaller apertures and camera passes from a higher distance.
“Krakatoa is an amazing manipulation toolkit for processing point cloud data, not only for what I’m doing here but also for recoloring, increasing density, projecting textures and relighting point clouds. I have tried virtually all major point cloud processing software, but Krakatoa saved my life on this project,” Kizny noted.
In addition to using Krakatoa to visualize all the CG components of Rebirth as well as render point clouds, Kizny also employed the software for advanced color manipulation. With two subsets of data – a master with good color representation and a target that lacked color information – Kizny used a Magma flow modifier and a comprehensive set of nodes to cast and spatially interpolate the color data from the master subset onto the target subset so that they blended seamlessly in the final dataset. Magma modifiers were also used for the color correction of the entire dataset prior to rendering, which allowed Kizny greater flexibility compared to trying to color correct the rendering itself. Using Krakatoa with Magma modifiers also provided Kizny with a comprehensive set of built-in nodes and scripting access.