Harry Potter and the Deathly Hallows: Part 2, the eighth and final film based on author JK Rowling’s famous series of books, is, by all accounts, one of the best, perhaps the best, movie in the franchise. A momentous and thundering final film. It is also largely digital, with 1800 visual effects shots created by several hundred artists, most of whom work in London-based visual effects studios. These artists extended sets, placed footage of actors filmed on greenscreen into digital backgrounds, added CG characters, removed the evil Voldemort’s nose once again, and, this time, destroyed a digital Hogwarts.
Senior visual effects supervisor Tim Burke, who had won an Oscar in 2000 for supervising the effects in
Gladiator, has been creating such magic at Hogwarts, the school where Harry Potter and his friends studied magic for 10 years. He joined the Potter crew on the second film,
Harry Potter and the Chamber of Secrets, which released in 2002, as a visual effects supervisor at The Mill. On the third and fourth films, he became the co-VFX supervisor for the postproduction efforts. He then took charge of the final four, working with director David Yates as the overall visual effects supervisor. Burke received an Oscar nomination for the third film,
Harry Potter and the Prisoner of Azkaban, in 2004, and the seventh,
Harry Potter and the Deathly Hallows: Part 1, in 2011.
During the past 10 years, dozens of visual effects studios have worked with Burke and the Potter directors, and as the series progressed, so, too, did the kinds of visual effects the filmmakers could achieve. Here, CGW contributing editor Barbara Robertson asked Burke about the ways in which this film differed from the previous.
When did you begin filming the final film, Harry Potter and the Deathly Hallows: Part 2?
We filmed Parts 1 and 2 together, starting in November 2009, and continued through the summer of 2010. But, we filmed out of chronological order. We did the battle sequences for Part 2 before we shot Part 1. To figure out how to coordinate all this, we did an intensive period of previs.
What do you mean by “intensive period of previs”?
We worked on previs for nine months. We didn’t have a finished script for the film. It was always evolving. We didn’t have a shooting script. We didn’t have storyboards. We had the book, the unfinished script, a brief from David [Yates], our previs assets, and four or five animators. So, we worked with David and designed the whole thing in previs.
How detailed was the previs?
We previs’d all the action sequences–the Death Eaters, the giants, the attack from the bridge, all the action until the giants breached the school. Ferran Domenech from MPC (The Moving Picture Company) led that effort. We developed the way things moved, the knights, the giants. We animated everything. We built and rigged additional assets as needed. We really were at the forefront of the ideas. It was a great, creative, fun thing to do.
Why did you do such extensive previs?
We developed all this previs in animation to help everyone understand the action and all the pieces of the puzzle. It was a guide for the live-action shoot, not just to run sequentially and fill in gaps. The design process had to happen early to decide what effects we’d need. It helped us determine which areas of the school to build. And, it gave the whole story of the battle; it described the battle sequence from start to finish. We ended up with 30 minutes of previs, complete edited footage. We gave it to David Yates and his editor, and he used this to cut with; the previs stayed in there as he shot. So, it became the film, basically. And it turned out to be pretty close to what we previs’d.
Hogwarts didn’t appear in Part 1, but the action centers on Hogwarts in Part 2. Did you do set extensions for the Hogwarts miniatures that were used for the previous films?
The action sequences we developed in the previs take place around the school and the hillside where Voldemort arrives and stages his initial attack. We didn’t have scripts when we started in 2008, but we had read the book. We knew we needed to create a flexible asset. We have always used our miniature model for Hogwarts, but we knew we would need to redesign it to encompass specific areas of action for the battle. We needed to add a huge viaduct on the front of the school, the main entrance where Voldemort would attack and where they would defend the school, the open courtyard, battlements, and the wooden bridge, which is a main attack route from one side of the school. None of that existed. So, we did away with miniatures.
Voldemort and his armies destroy a digital Hogwarts within a digital environment created by
artists at Double Negative and The Moving Picture Company. He lost his nose, though, at Cinesite. |
Who built the digital Hogwarts? Did they build a CG version of the entire school inside and out?
Double Negative started the building and destruction process for Hogwarts in 2008 with Stuart Craig [production designer] while we were still filming
Half-Blood Prince; it took 18 months. Because we had designed the sequence in previs, we realized we didn’t need to build the back half of the school, so we focused on a 180-degree area. Stuart Craig designed an iconic image of what the partially destroyed Hogwarts would look like. We filmed on partial sets and greenscreen—we had a partial set for the courtyard, a separate set for part of the viaduct and part of the bridge, and an interior set for part of the battlement.
We take the camera from the far hillside in one move, through a window to find the Death Eaters fighting. Voldemort’s army arrives on Voldemort’s rock on one side, and on the other side, the snatchers (the humans who turned to the dark side) attack on the other bridge. One of the biggest jobs was linking all these action sequences and the environment together. David wanted signature camera moves.
What do you mean by “linking”?
When I say ‘linking,’ I mean visually putting the world together so the audience understands the layout even though these things were fairly small sets with greenscreen. Because of the shot design, we see areas in many shots that obviously didn’t exist. Sometimes we had a real set in the courtyard and everything else was CG. We had elements of the viaduct and everything else was CG. Even if the set is prominent, everything behind it is CG. We built the rock Voldemort arrives on; it was less than 50 feet across. Then we put that rock in the middle of a [CG] environment, put thousands of Death Eaters around him, and had huge, swooping camera moves. We created the geography of the world and dropped it around the sets using a correct combination of different pieces. It was a big jigsaw puzzle.
Had you done environments this extensive for the previous films?
The biggest environment in a prior film was the opening sequence for
Half-Blood Prince, which was a rebuild of London from Trafalgar Square to Charing Cross, also by DNeg. But, this was more ambitious. It covered a far greater area, and the resolution was far greater.
Our environments have gotten better and better with every film. We surveyed environments in Scotland, and we have the topography, so we no longer sent film crews to shoot plates. We created a Scotland-like world that has a loch, mountains, the school, and several miles of environments that we built and worked over. We see it in the daytime, sunlight, predawn, and at night—all these lighting conditions. It was lit and relit, and it looks photographically realistic. When you watch this film, you believe it’s there. The environments are not off in the distance. We placed the action in the middle of them.
In terms of pushing the technology, this is one of the biggest things. We couldn’t have done it a few years ago because of the sheer amount of data. I think the technical assets and software developed to handle all this are phenomenal and don’t get much appreciation. People take environments for granted; they don’t think, ‘Hold on. None of this is real.’ But, they gave us flexibility.
More flexibility than when you worked with the miniatures on set?
Before, when the school was a miniature,we’d shoot the background plates, and then months down the line, we wouldn’t be able to change anything. On this film, we redesigned the whole end sequence with Voldemort four weeks before we delivered the film. David [Yates] was putting new ideas in, and because we had this huge digital asset, we could do them. This would have scared the facilities a few years ago. But they did such a great job creating the assets that we could keep changing the shots. We had great flexibility at the back end. I think it is one of our biggest achievements.
How did you allocate the work?
We always knew that no one facility could take all the work of the battle, so we had to split the action cleverly. We gave everything on one side of the school—the end of the courtyard and the beginning of the viaduct and the hillside—to MPC because they could use their crowd software, Alice, for the Death Eaters and the crowds. Then, because the giants and knights featured heavily in the battle, we gave that work to their animation team.
MPC handled the main action, the knights, the giants, Voldemort’s army for the attack, and the fighting. They also did the fire simulation—they created animals out of fire. For the giants, we decided to use actors and scale them and have CG facial prosthetics for the final close-ups. Everything on the other side—the bridge, the destruction of the school, and the aerial work—went to Double Negative; DNeg did the environment work for the school and bridge destruction.
So, one became the miniature for the other. MPC would do the animation and work out the action, and then provide cameras to DNeg to create the background, light it, and composite it. Or, if DNeg did the action in the foreground, MPC would do the background. These two vendors had the heaviest work.
It sounds complicated.
It gets further complicated because we delivered a lot of these shots as full-stereo shots.
So, we developed a pipeline along the way to share stereo rigs and cameras between facilities.
The hero performance would set the stereo for the camera and pass it to the background facility,
which would render left and right versions. Everyone worked on mono versions, so we had three things running at the same time for over 200 of the shots. It was a massive logistical job as well as a great technical exercise.
| Animators at Double Negative freed an albino dragon, created at the studio, from its imprisonment and taught the tortured animal how to fly. |
Did the addition of stereo 3D to this film affect the design?
It did, actually. David’s take was that [stereo 3D] should be immersive and not take you out of the picture. He shot it in 2D, and it has a lot of drama and action. He was keen that [stereo 3D] didn’t distract from the drama, so we kept it subtle and all about the actors. The volume is in the actors, not in the space, and I think that works very well. We tried to exploit the depth without making people feel uncomfortable; we’d get creative with the stereo and then back off for the dramatic scenes.
In the big linking shots where we establish parts of the school, the audience feels like they’re flying around watching the action below. These flying shots are incredibly spatial. And, we have a roller-coaster cart ride in the vault; we exploited that. Also, we deliberately had the dragon break the screen plane, and had some real fun with that.
Who worked on the dragon?
DNeg. They did a great job with the dragon in the Gringotts’ cave. We developed a whole back-story of the life of the dragon. He’s a semialbino dragon that has been abused and kept locked up in the bank. The [Gringott] goblins torture him so they can access the vaults. We referenced badly treated dogs. The poor animals. Their faces and bodies displayed the level of emotion we wanted. And then we designed a badly maltreated creature whose posture and body poses convey that he’s been tortured. He’s never seen the light of day. We see him early on in the film. When he’s set free from his shackles, he sniffs the air, follows the scent of fresh air, and scales the walls with the kids on his back. He gets revenge on the goblins, and then he discovers how to fly. We have an animated sequence that shows what a dragon looks like when it can’t fly properly. The sequence is all-CG; there are no miniatures. He’s our biggest animated character, and he’s completely unique. But, of course, we had lots of other things–Dementors, Death Eaters, all the usual CG doubles. . . .
Which studios worked on the other characters in the film?
Rising Sun, which had done the Dementors in Part 1, did all the Dementors in the action sequence around the school. Baseblack did the ghost, the gray lady, and they also handled Scottish environments for the sequences we shot on the back lot.
Cinesite came on board to remove Voldemort’s nose and do a revamp on the marble staircase. Stuart Craig wanted a more intricate series of zigzagging staircases that were wider to allow for the big fight sequences. Their work involved massive set extensions in both directions for one of our big interior sets at Pinewood. Cinesite did all the work on that, plus the physical dynamic destruction, wand effects, and crowd replication to populate it.
Framestore did the Basilisk [the giant snake] and the Chamber of Secrets, which was a set in the second film. They modeled, textured, and created the environment by referencing film clips from that film. They created a nice water simulation—a wave of water which attacks the kids that we shot on greenscreen.
Framestore also did the Kings Cross Sequence, which we call the white mist. This is the heaven sequence where Harry meets Dumbledore. It’s a white, ethereal version of King’s Cross station. And then they did another version of King’s Cross for the very end of the movie. The
original shoot took place in the real King’s Cross, but they did a re-shoot in Levesdon.
How many studios worked on the film?
We had 11 vendors, including several small vendors that helped us out with shots, and Vine, our own little in-house facility where two or three people working on 2D systems ended up doing a couple hundred shots on both the final films. Lola, which is famous for youthenizing actors, helped make Alan Rickman 30 years younger for flashback sequences, and they also had a call late in the day to do aging work on the 19-year-old actors at the end of the film. They treated all those shots digitally to make the kids older.
And then Tippett Studios came in late to help us finish off a big technical animation exercise involving a fun sequence where the kids break into a vault and try to find the horcrux. There’s a spell that causes the treasure to multiply, so they created a dynamic simulation of the multiplying treasure. As the treasure multiplies, it fills up the room and elevates Harry on a growing mound to the point where he can reach the cup. At the same time, everyone else is drowning in treasure. It’s a great little sequence.
Gradient Effects in Berlin and Marina del Rey [California] did fluid work on sequences for some flashback memories. They’re very good, very technical. Union VFX in Soho, which recently did the work on Danny Boyle’s 127 Hours, did 50-odd shots at the end. And then, of course, we had extra vendors to do the stereo conversions. We were flat out for the last two months, but it all got done.
What will you do next?
I think it will be a big effects film based in the UK but using global facilities. One thing we really got working on Part 2 was treating the whole of Soho as one big facility. We shared work among four and five vendors, painlessly moving between. It’s like having a huge resource at your disposal. They all worked together creatively rather than competitively. [The next project] is a bit of a secret, but if it happens, it will be exciting