The project began with extensive research into the city of Shanghai as it existed in 1945 for John Woo's epic film "The Crossing". In addition to the package of historical photos we received from the client we poured over websites & historical archives to build our own collection of images. These images became the base of everything we did on the project, we used them as reference for everything from the layout design, street planning and texturing to final detailing and arrangements.
My involvement at this early point was centered around initial tests about what technology we would use to build and render the shot and creating rough layouts and camera moves that we used to previs and plan the shot. It became immediately clear that this shot would require a fully 3d build on pretty much everything, there really was no way to fudge or rely on tricks because of the extreme and dramatic camera move.
The base of the build was the street map layout. This was created using Esri's cityEngine and was based directly on maps of the city as it existed in the 1940's. Though we fudged some streets nearer the camera to get a better look, and the streets in the distance beyond where we can recognize actual street placement was completely imagined, most of the streets represent exactly the layout of Shanghai.
This became the "Master Layout" and we then divided the city into sections - and this was the plan also for how the layout would come together - each artist who would be working on the shot would be given a different section and each of these was to be fed back into the Master Layout. This would allow us to manage each section separately but in unison with the others and would allow the whole city to progress at the same time.
The asset build was a sort of two-tiered effort, a combination of creating a mass of "generic" buildings that would used for filling in the large part of the city, and the "Hero" assets that would represent the famously identifiable & recognizable buildings of the historic "Bund" area nearer the river. And in addition to this there was also a need for tons of little "filler" assets such as road signs, street lamps, signage, trash cans, cars, ships... all of the items that bring life and realism to a shot like this. All in all we built & gathered and textured about 360 separate assets. Much of the generic asset builds were done by me over a three week period before any layout work began, and we had a couple model and texture artists working on the hero assets. All in all we built over 350 separate assets for the shot.
The Crossing is a two-part epic disaster movie directed by John Woo (Red Cliff, Mission: Impossible II, Face/Off), set during the Chinese Civil War, and climaxing with the sinking of the luxury liner Taiping in 1949. In the first film, multiple plotlines track the interweaving fortunes of a number of characters through romance and war, setting them up to converge with catastrophe in the closing instalment. Part 1 was released in China in December 2014, with Part 2 due for release in May 2015.
Tippett Studio contributed a number of visual effects shots to The Crossing: Part 1, most of them involving large-scale environments. Key sequences include a re-creation of the bustling city of Shanghai, scenes of the Taiping at sea, and a dramatic battlefield flypast.
In this exclusive VFX Q&A, Tippett Studio’s environment art director, Kent Matheson, reveals the painstaking process of bringing to life the historic events revisited by Woo in the film that’s been described by some as “the Chinese Titanic”.
How did you get involved with The Crossing?
Tippett Studio had worked with John Woo on Red Cliff, so we were on his radar. The connection for The Crossing was made by his post-production supervisor in Beijing, Andy Chen. We had an initial kick-off with John before he was consumed with all the other facets of the film – after this we worked through Andy, who would show our work to John and get his notes back to us.
How many shots did you deliver, and who was on the team?
We delivered ten shots during a two-and-a-half month period. It was a very tight schedule in which we had to research, develop, and execute the shots somewhat simultaneously.
We had a crew of about 20 on the project, divided into separate teams for each environment. The visual effects supervisor was Chris Morley, with Ken Kokka and Yimi Tong producing. On the big Shanghai shot, we had Ross Nakamura as the lead compositor, and Chris Paizis was our head layout guy and scripts wizard. Ben Von Zastrow was our lead technical artist, and Darin Hilton and Pamela Saad contributed immensely.
Let’s talk about the big environment shots. What did they comprise?
There’s a shot that lifts up from the street level of Shanghai’s slum area and travels completely across the city towards the wealthier area of the Bund. It’s kind of a signature image for the film. It describes the contrast between the different levels of the culture – the extreme disparity between the wealthy and the poor areas.
We also created a series of shots of the ship, Taiping, sailing on the open ocean, using plates that were filmed in a studio backlot. In one shot, the camera follows several seagulls as they traverse the length of the ship, passing over the crew and passengers on the various decks before reaching the lead character standing near the prow of the ship. The shot involved water simulations, atmospherics, a full CG ship blended into the plate, and a whole lot of fancy comp work.
In another shot, we explode out from near the barrel of a large cannon to travel over an enormous battlefield full of advancing soldiers and tanks. This began as an aerial plate showing only a few soldiers and a single tank; we filled the land with tanks and CG soldiers, and filled the sky with exploding shells.
Watch Tippett Studio’s video breakdown of the Shanghai establishing shot:
How did you go about researching 1940s Shanghai for the big establishing shot of the city?
Shanghai at the time was an amazing mix of cultures and styles, with aspects of the old world and new mixed together, so we found photos that really captured it well. Hooray for the amazing resource of what people share on the internet!
We collected everything we could – not just the obvious buildings on the Bund, but also old signage, dock photos, posters, people, cars, boats, bicycles, streetlamps, flags and clothing. We wanted to have a very solid idea of each district of the city, so that all the parts we created would properly support the final look and be accurate to the place and time.
Aerial views gave us a sense of the textures and densities of the different regions of the city, while street-level images showed us what we needed to properly detail it. In some cases, we were able to use the reference to directly inform the shot creation. For example, we projected an archival image of the Wusong River on to our CG river plane, and dressed in the boats and supplies in the positions shown on the picture. Chris Morley loved the natural chaos of the photograph, and didn’t want to lose it. You can’t make that stuff up!
How did you begin laying out your CG model of the city?
One of the things we had found in our research was original maps of Shanghai from 1940, so we based our initial layout work on these. We located the key areas needed for our shot, and created a very simple representation of the city, laying in the rivers and bridges in the correct places and using blocks for city buildings and simple models gathered from SketchUp for the landmarks.
We also blocked in the camera move at this stage. We explored various paths, ending up with a move that pulls up and over the secondary river in Shanghai, glides over the commercial district toward the Bund, and finally drops into an upper window of the Peace Hotel.
What was the next stage, once the layout was approved?
We used Esri CityEngine to lay out the streets. We built only what would be seen in the shot, matching to the maps but also fudging a bit where we wanted to create dramatic focuses. This became the base on which we placed the buildings and details.
How did you create the various assets needed to bring the layout to life?
We had open categories for assets such as “boats”, “cars”, “trucks” “props” and so on. We divided the buildings into different types – “residential”, “commercial”, “industrial” – and had different styles within those, from older wood structures to more modern ones. These were set up in different configurations and groupings, from single buildings to whole neighbourhood blocks. At first we used a fairly random scattershot approach, but we became more focused as the categories filled up.
In all, we ended up with around 375 separate assets feeding into the shot. All of these were created as fully 3D models, viewable and renderable from all angles. But we were careful to control the level of detail – we built and laid everything out it very much with the mentality of a matte painter. And the requirements of the shot meant that none of the assets needed to be built to what we would usually consider a “hero” level of detail.
What software did you use to build the models?
Many of the models began life in SketchUp, but all were cleaned up, textured and shaded in Maya. Every building and prop was set up with the same mapping inputs, and many shared shader settings – this allowed us to automate and regulate many of the processing tasks.
With so many objects to bring together, we decided to radically simplify the UV and texture process. Therefore, almost all of the objects were baked down to a single UV set. We decided on four basic shader inputs, and then used a single map to drive each of these. This let us speed up not only the creation process, but also the rendering times and memory usage further along the line.
With the models constructed, how did you begin assembling the shot?
We were handed the live-action plates of the head and tail shots – the “bookends” – and given a verbal description of the shot, along with a rough idea of the timing John Woo had in mind.
We broke up the city into different sections and subsections, each of which could be worked on and published separately. Anyone on the team could jump in and work on a specific area as needed, so we were able to “multi-thread” the layout and population effort. We had a few hitches from time to time – cars driving over people or people walking into buildings – but all in all it worked out well.
We added and animated cars and people into the streets as one of the very last stages. The people were realised via very simple models, pre-animated with various walk and talk cycles before being placed into the scene. Additional life was added with flags and other minor animated pieces.
Describe the process of rendering out the scene.
All the separate city sections were collected into a master scene file, and the final rendering was done using V-Ray. We made some touch-ups using Glyph’s Mattepainting Toolkit, but most of the scene’s look and detail was handled via the raw CG assets. We also relied heavily on lighting and atmosphere.
The renders were split into various passes and assembled using Nuke. Early on in the process, we had identified several key frames that best represented the major elements of the shot; using these, we’d made well-defined key art representing what Chris Morley wanted to see. That was what we presented to John Woo.
We used the key art throughout the process for reference to lighting and detailing, but it was particularly useful in the comp stage as the rendered layers were assembled. What really helped was having a strong vision of what the shot was going to look like and stickling to it. Chris’s background is in compositing, so he had a strong feeling for how he wanted the elements to come together.
Were any significant changes made to the shot as your work progressed?
We could have adjusted quite easily to camera changes, since we’d built our assets to be used wherever we needed them. But it wasn’t necessary.
However, John Woo liked what was happening in the shot so much that he asked us to extend its duration by about 15 percent. We loved this, because the slower camera let us see more of our work!
Tippett Studio is strongly associated with character and creature work. Will you be doing more of these big environments in the future?
We are known for our rich history of animated creature work, but we’ve been doing environment work going back to The Matrix Revolutions, Constantine and Starship Troopers. However, that’s always viewed secondary to our creature work outside the studio.
During the last five years, we’ve really amped up our desire to accommodate more environmental work, and improved our pipeline accordingly. We have strong artistic vision and processes in place that allow us to bring large shots together quickly and flexibly, and we’re definitely looking to build on that more and more. The Shanghai flyover shot was a fantastic opportunity to prove that Tippett can quickly pull off crafted, large-scale environments without a creature in sight!
Currently we have a really exciting project in-house that involves flying over various incredibly dramatic landscapes, cloudscapes and cityscapes. All in all, it’s about five or six minutes of footage. It’s really fun stuff!
Did you draw any lessons from your work on The Crossing?
It was a whirlwind of an effort, and a lot of work came together very quickly. Some things we would do differently, of course, mostly in the area of what tech we would choose. The render tech we used on The Crossing was V-Ray and while it worked out incredibly well for us we’re currently looking into a program called Clarisse, which functions a lot like Katana, but with many more bells and whistles for environment work. It has everyone very excited.
Overall, I think everyone was really pleased with how smoothly the team worked together. We have experienced people in place who respect each other’s involvement, and I think we made all the right choices for the time and effort.
- Tippett Studio
- Kent Matheson at IMDb
- Chris Morley at IMDb
- Andy Chen at IMDb
- Clarisse – Isotropix
- Esri CityEngine
- Mattepainting Toolkit – Glyph
- Nuke – The Foundry
- V-Ray – Chaos Group
Special thanks to Niketa Roman.