Gramercy Park Studios Takes Flight through the Louvre Museum

The artist takes the viewer on a tour of cityscapes and creates a sense of moving images through the use of the light, reflections and textures he finds on glass, concrete, spiral staircases, paintings and people. The camera work is expressive and distinctive, and shifts from nearly static to observational to much more dynamic, sometimes using trompe l'oeil. The scenes were shot in Paris, Sao Paulo and Toronto, capturing locations ranging from the historic Louvre Museum to the modernist buildings of Oscar Niemeyer in Brazil and Mies van der Rohe in Canada.

Sequences at the Louvre, in particular, elaborate on the possibilities of camera moves, including swooping, twisting and flying, to give the museum character and storytelling potential. Gramercy Park Studios contributed unusual VFX and point cloud imagery to enhance the scenes with elegant, extreme camera work.

New Perspectives

“Recreating the famous Louvre Museum was a challenging and engaging project from both an artistic and technical point of view,” said Francisco Lima, VFX Technology Supervisor at Gramercy Park Studios. “The beautiful results we achieved meant pushing the boundaries of our workflow to achieve new dimensions and angles the director wanted, within the digital environment. Watching ‘Invention’ really does make viewers feel like they’re inside the Louvre, and experience it from a perspective that would not be possible otherwise.”

To achieve Mark’s vision of blending a live action shoot and virtual environments, Gramercy Park Studios’ work involved techniques in image acquisition, geometry extraction, 3D modelling and digital paint. "The initial step was technical previs, based on architectural plans of the Louvre and photographs from a location scout Mark and our team undertook, walking around the areas that he was thinking of using for his film,” said Francisco.

"We created a low resolution, version of several rooms in the museum, optimised for speed and ease of use, that Mark could use to navigate virtually through the spaces and decide how to best shoot the project. We have a special CG camera rig built in Maya that the CGI team uses for previs sessions with clients. Using the optimised model and this rig, Mark could quickly design and set up his camera moves, and play them back in real time,” said Francisco.

Virtual Shooting

As well as movement and speed, the previs helped him choose which camera rigs and lenses to deploy during the live action shoot. “He decided on the RED Epic early on, so we were able to configure our previs camera to match the RED Epic. That way, whatever lens we selected in previs, would match what he would see when shooting the live action.

“Mark visited the Louvre several times during this stage, and after each visit he would get a better idea of the areas he’d like to portray in his film and how he would like to capture them. Then we would do a new previs session so he could put his ideas in action and see how they would work, in effect doing ‘virtual shoots’. The DP would also come to some of the sessions to see the type of move required, exchange ideas on how to capture the live action, and start getting a list of equipment ready for the shoot.” From there, with the physical limits defined, Mark could also determine which rooms to build virtually, allowing them to create the extreme camera moves seen in the final film.

Inside the Louvre

During the next phase, the Gramercy Park team visited Paris several times to carry out a thorough photographic capture of the two selected interiors. “We planned a one-day shoot in each room, and used several Canon 5D MkIII cameras for the task,” said Francisco. “Because we were going to build the ceiling of the second room entirely in CGI, Aviv Yaron joined us and conducted a special photoshoot of the ceiling with panoramic equipment, including a panoramic head. Yaron is one of the most experienced VFX/CGI photographers I’ve ever met and works with his own gear. 

“We had a couple of meetings prior to shooting each environment. During these meetings we would watch the previs clips Mark had defined, from which we would create a shot plan to guarantee that, as far as possible, we would be able to photograph the areas the previs camera was seeing. At the same time, we took specific measurements throughout the photoshoot. Allied to the photos, these measurements were used to generate an accurate point cloud at the correct scale. We also decided early on that the light of the scenes should match the existing light at the museum, and kept the light captured in the photos on the day of the shoot.”

Calibration & Point Cloud Generation

From the photography, high resolution textures and geometry were extracted using a photogrammetry workflow in which, back at Gramercy Park Studios’ London facility, the hundreds of photos were calibrated within the virtual space used Photoscan from Agisoft. This involved finding the exact position in 3D space from which each photo was taken in the Louvre. From this extensive calibration, the team were able to extract a base geometry and point cloud from which detailed and precise 3D models were created. Extracted digital textures were then mapped, edited, digitally enhanced and re-projected onto the 3D models.

Francisco said, “The final point clouds for the winged victory environment contained the entire building structure, walls, floor and ceiling and the beautiful winged victory sculpture itself. The structure point cloud was used as reference to model walls, columns and so on, while the point cloud of the statue was brought into Mudbox and used as a base to generate a proper, production-ready 3D mesh.

“The second room, which we re-created in CG, has an amazing, detailed ceiling. the same process as the one we used for the statue was used to, again, create a production-ready 3D mesh for the ceiling details and the eagles. The images here show all of the calibrated cameras and the resulting 3D mesh, and a textured version. It’s not the final version, but it gives a clear idea of the quality we were able to get from the excellent photos that Aviv had taken.”

Matching the Live Action

Following sculpting in Mudbox, their CG pipeline consisted of modelling in Maya and texturing with MARI, rendered in V-Ray. Compositing was completed in Nuke X. To make the transition between the live action footage and the digital environment invisible, Gramercy Park blended the real and the virtual.

Based on the geometry of the interior they had captured and digitised, the team recreated the camera move featured in the live action footage digitally in order to blend it with the digital camera, approaching the transition as a matchmove using PFTrack. The result is balanced between both environments and accurately portrays the kind of voyage Mark wanted to take inside these detailed spaces.

The film ‘Invention’ made its world debut at the Toronto Film Festival in September, and has also been shown at the BFI London Film Festival.  www.gramercyparkstudios.com

Words: Adriene Hurst
Images: Courtesy of Gramercy Park Studios