DNeg is known for its Oscar winning feature film work, but it also has a strong television group, founded four years ago. The TV team started with a crew of seven artists and today is 170 strong. It currently has 8 shows in production, and it will complete about 20 by year’s end. The TV team operates from three sites, London, LA and Mumbi and services customers such as HBO, Sky, Netflix, BBC, TNT and ITV.
Mark Pascoe is now Head of 3D and CG supervisor for the group. We caught up with him recently to discuss the style of work the team does. Pascoe partially jokes that the TV team does 90% of the work of the feature film group in 10% of the time. Pascoe had been a strong RenderMan user until about 2014, but under the weight of TV timelines he has become an ardent user of Clarisse. So much so, that now every show starting at DNeg has Clarisse as its primary Renderer. Given the speed and nature of the work, even with the variety of things the team takes on, a shot may not leave the desk of a particular artist, as the TV unit is staffed by primarily with generalists and hence one person can take a shot right through the pipeline process.
The TV division is a separate department within Double Negative, but is able to lean on the robust pipeline already in place, settling on a strong combination of Maya, DNeg’s in house tools, such as Riot, the Foundry’s Nuke and Clarisse 3D for all the TV work.
As the Head of 3D, Pascoe enjoys the physically plausible approach the DNeg takes to all its TV work, and he finds Clarisse provides four key things that sold him on a pipeline centered on that software.
- The speed of rendering
- The customisable templates
- The shallow learning curve for new recruits
- The scalable nature of the tools.
Pascoe uses Clarisse as much more than just a renderer, “we actually use Clarisse as a layout tool too”, he points out. Solving complex lensing and blocking problems such as they did in The Young Pope.
The Young Pope
The Young Pope, a project that used Clarisse for it’s complex 3D visual effects. Pascoe pointed to it as a great example of their workflow, the show required 900 shots over 6 months with the program first airing late last year.
Jude Law stars in the Young Pope, a European drama based on the story of the first American Pope who takes the name of Pius XIII. Shortly after his election, he causes upheaval in the Vatican with his desire to reject both publicity and the liberal philosophy of his predecessor. Pope Pius XIII was former Archbishop of New York. The film also stars Diane Keaton as Sister Mary. The production required not only set extensions and Vatican environments, interior and exterior, but it also required such unusual elements as a pyramid of human babies, and vast complex crowds.
The project was financed by Sky, Canal+ and HBO, and visual effects were all by DNegTV. The series consisted of 10 episodes with a budget of about $45 million. The project was shot on the Red Epic Camera in the famous Cinecitta Studios of Italy, but locations such as the Sistine Chapel and Papal Library were not available as an option for main unit to film.
While the team did the exterior of St Mark’s Basilica and other key landmark sets, the visual effects for the show actually started with a giant pile of babies that the camera pans over before the young Pontiff climbs out from the pile and walks into Piazza San Marco commonly known as St Mark’s Square.
The later half of the shot was filmed with a different lens from the start, to give the director flexibility to re-rack in post, which was both heavily CG and composited live action. The shot is complex. Not only as it is some 1000 frames long, but to solve the blending the team ended up unusually “animating the actual film back, which is very rare” Pascoe explained.
The final shot is a combination of live action, CG babies and live action and set extension. To do the digital babies Dneg’s Leonardo Vitti came up with a scheme of building the moving mass of sleeping babies as a series of baby triangle patches. The babies were all hand animated but a core cluster of 34 babies was built in a triangular formation. This patch was then interlocked with instanced versions of the same patch 16 times for the illusion of 534 babies. This was started in Maya and then moved to Clarisse for scene assembly and shot finishing, with Vitti also developing the Lookdev.
The shot was so successful that the production decided to revisit the pile of babies in a later episode, with a similar setup, but half the duration. Unfortunately it was required to be shot from a completely different direction on the babies.
In order to assess the scale of the task at hand and help define shot workflows, the team scattered generic baby poses, almost like previs, in order to get a starting point. “I really like to encourage versions, – I like to have something to work from”, explains Pascoe. by building up the shot this way, the team could look at the 534 babies and pick which needed to be additionally hand animated. For close up areas where there was a need for specific animation, the team could isolate and replace say baby #6, and #17 from a triangle patch and replace just those babies with new hand crafted animation. In the end about 95% of the babies are animated with a neutral loop and 5% had to interact with the tracked live action.
For the massive crowd sequences the team faced a difficult problem. The Vatican does not allow film crews to film and survey as one can do on a traditional set. But it does welcome tourists!) Armed only with hand held cameras, the DNeg team went to Vatican City and took some 1200 tourist photos, all without a tripod, so HDRs were impossible. While this got some great reference, there were some unexpected aspects. While most crews build up when doing set extensions. DNeg had great source looking up, but the actual floor of the Vatican was nearly always covered by the crowds of tourist. Hence the team got “very little below 10 feet off the floor” joked Pascoe.
There were some crane shots in the final sequence; these were all exteriors and all wide big establishing shots. Photogrammetry was key to matching live action to these plates. On the day of the crane shots, the sky was overcast which was perfect for the 3D team in terms of diffused light and no sharp shadows, but the plate photography was naturally flat. The team rebuilt the scene in 3D, not to replace it but use the tracked 3D to relight the live action. The 3D matched geometry was so exact that the plate photography could be relit using shadow passes and mattes from Pascoe’s team.
The crowds themselves were generated in DNeg’s Riot in house software in their Mobs format. This had been written for RenderMan but it was modified for Clarisse. The CGI agents, many with umbrellas for the rain shots, were built off a set of motions captured in a Moven Motion Capture Suit and feed into DNeg’s crowd simulator. The team used Clarisse to instance agents onto scatterers as a vehicle to get quick layout variations for director approval. “The plan was always to revisit these layouts and rebuild them using their crowd simulation pipeline, however the result from Clarisse was at a high enough standard to service the background crowd in full shot production, meaning it was only ever the foreground patches that were simulated (patches A and B below), reducing the time needed for traditional crowd work” explains Pascoe.
The vast crowds posed a unique problem. The ground plane in such a vast square would normally be flat, and while walking the space in Italy, the team felt that it was level. In reality, the Vatican square has a 3 degree slope that allows the real crowds to see the real Pope more easily. “The shape is actually more of a bowl than even a perfectly flat or sloped floor plane,” Pascoe explains. DNeg had build very accurate models and populated it with 100,000 agents, but it would not match the plates until the team stopped bug hunting their 3D and worked out that real environment had been so cleverly designed.
The ability to quickly work out shots proved invaluable for Pascoe, who says that “this ‘throw anything at Layout’ turned out to be key to getting the finals”. The team needed to work quickly and “by being able to change the underlying geo – change the layouts, then by scattering have lose agents on the right plane really quickly was brilliant,” he relates. Pascoe was especially impressed with how much they could render in real time and thus how quickly they could iterate on a shot to get what was needed.
The team worked on several cities sets. In one dramatic shot, the team was required to complete a pull out from Venice to outer space, which required Geo for all of Venice. For the primary exterior location the team was able to fly an Octocopter with a Red Epic Camera up looking over to the St Mark’s Square, and surrounding buildings. This was shot at about 7 to 8 in the morning to a height of about 350m. This provided again excellent reference, and some good plate shots. The Octocopter footage provided invaluable Photogrammetry information, which served as the foundation for the layout.
The team also had to do a lot of green screen work, and some unusual isolated animations such as a 3D flower opening and matched into a live action plant.