One of the biggest Chinese films in recent years is Zhong Kui: Snow Girl and the Dark Crystal, an epic fantasy that reimagines China’s legendary folk hero, Zhong Kui, as a shapeshifting superhero with a Hulk-like inner self.
Co-directed by Peter Pau, Academy Award-winning cinematographer of Crouching Tiger, Hidden Dragon, and Zhao Tianyu, the film chronicles Zhong Kui’s (Chen Kun) journey of self-discovery as he battles to prevent the demons of hell from invading the world of mortals, while attempting to romance the porcelain-skinned Snow Girl (Li Bingbing).
Most of the 1,200-plus visual effects shots in Zhong Kui: Snow Girl and the Dark Crystal were created by Macrograph, a Korean VFX company, with director Pau fulfilling the role of overall production VFX supervisor. However, for a key action sequence near the movie’s climax, the filmmakers turned to Pixomondo’s Beijing facility.
In this exclusive Q&A session, Cinefex spoke to Pixomondo’s VFX supervisor, Wil Manning, who described the challenges involved with delivering a full-throttle, full-CG sequence to a tight budget and in record-breaking time.
Watch this detailed breakdown video of Pixomondo’s visual effects work for Zhong Kui: Snow Girl and the Dark Crystal:
How did you get involved with Zhong Kui?
One day, in August 2014, I was pulled into the Beijing Pixomondo conference room by our management team. They asked if I would drop what I was doing and jump on to this sequence for a Chinese fantasy film called Zhong Kui. I had no idea what they were talking about.
John Dietz, our former head of production, told me, “Zhong Kui is this gigantic possessed demon hunter, and he fights the even more gigantic Demon King, who has a flaming head and snakes sticking out of his back. They fight through a village, destroying it as they go, then Zhong Kui lures the Demon King into a lake, where they fight through ancient sunken ruins until the Demon King causes a volcanic eruption that levels the underwater environment. Zhong Kui tricks the Demon King, cuts off his hands with a sword he pulls from his spine, and throws him back into the village. The sequence is three minutes long, stereo, full CG. You’ve got four months to deliver – and that time started two weeks ago!”
How did you react?
I blinked a few times, then asked politely if this was some kind of strange hazing ritual to the rookie VFX supervisor. As it turned out, the brief pretty much described exactly what I ended up working on for the next 14 weeks.
How did you set about tackling the enormous workload?
Our Beijing facility was very busy delivering Jiang Wen’s Gone with the Bullets at the time, so we had to reach out to what we affectionately term the “Pixoverse” for help.
My colleagues around the world were very patient in listening to my pleas for assistance, and only a few of them questioned the state of my sanity, given my requests. To be fair, this was expected – I think it’s really hard for people with a predominantly Hollywood background to understand this kind of Chinese brief. Fortunately, we’re a company full of brave and intrepid individuals, so help was forthcoming in the form of three teams spread around the globe.
How did you divide up the work?
We had a team in L.A. that handled previs, animation, look-dev for the characters, and the finishing of the village shots. Our team in Stuttgart, Germany, did all the heavy lifting for the underwater shots. And we had a small home team in Beijing that handled storyboards, design tasks and much of the TD setup work.
In total, we were 35 people spread over three time-zones. Key awesome people involved were Thilo Ewers (division VFX supervisor) and Sebastian Meszmann (VFX producer) in Stuttgart, Tim Jones (division VFX supervisor) and Julia Neighly (VFX producer) in Los Angeles, and Charlie Winter (CG supervisor) and Cinzia Wang in Beijing. I wore two hats as overall VFX supervisor and producer, which may or may not be responsible for the permanent facial twitch I carry to this day.
I think Thilo Ewers and Sebastian Meszmann are as responsible for the success of the underwater sequence as anyone, and I’m very grateful to them for their leadership and professionalism. I tried very hard to give them as much creative control over their own sequences as possible, and it was great that this was rewarded with strong results.
How closely did you liaise with the rest of the production?
On my second day on the show, I went to meet Peter Pau, who was director, producer, production VFX supervisor and cinematographer on the film. Peter is an accomplished and amazing veteran – a man who can wear as many hats as he likes without missing a beat.
Peter explained to us that the rest of the movie’s visual effects – over 1,200 shots – were being handled by Macrograph, and that our little sequence would slip in somewhere near the end of the film. Macrograph would supply us with the creature models, while Peter would supply us with lots of reference footage of the village. There were some storyboards, he told me, but we were free to throw them out.
The original character designs came from Weta. Macrograph created the character assets as Maya rigs with V-Ray for Maya shading and Yeti grooms. Everything else we developed for ourselves.
What was it like working with Peter Pau?
One of my favourite things about working in China is the direct access we tend to have to directors and key decision-makers. This show was no exception. Peter was always available to give feedback, and had an amazing ability to respond to emails with a comprehensive answer within twenty minutes of them being sent. He would do this day or night, from anywhere around the globe. I tried to keep up, but was shamed by my need for regular sleep. This kind of access to Peter, along with his rapid feedback, was one of the reasons we were able to complete the show in such a short time.
How did you plan the sequence?
The previs took five weeks. The animators built the sets as they went, and these were then exported and refined as they worked, eventually being updated with pieces of our kit-bashed environments. We sort of threw it all together at once – there wasn’t much time for anything else.
Did you go to L.A. to brief the animators?
Yes. I tied everyone to their chairs, and then forced them to watch my favourite kung fu action sequences. I talked about camera moves, quick takes and punch-ins, exposition of fighting and environment – a sort of crash course in Chinese “wuxia” cinema.
Which films did you show them, specifically?
One of the films we watched a few times was Crouching Tiger, Hidden Dragon. Peter won an Academy Award for Best Cinematography for that. There’s an action sequence early in the film where the two female leads battle it out over rooftops during the night. That was great, because it was similar to what we were going to do … just with fewer giant demons bursting through the walls and no volcanic eruptions. Still, I think any excuse to watch Ziyi Zhang fight Michelle Yeoh is entirely valid, and frankly action doesn’t get much better.
What lessons did you draw from that?
The shot count in that rooftop scene is really low. Instead of quick cuts, you’ve got long, following shots with a lot of great expository camera movement. One of my missions was to reduce our shot count, and watching Crouching Tiger, Hidden Dragon gave me a lot of ideas of how to do that. Eventually we got down to 34 shots.
Apart from choreography, what else did the previs help you with?
We made a conscious effort during previs to simplify the FX problems. When you have an average shot length of six seconds, and only five weeks for a massive amount of FX tasks, there’s a certain amount of danger involved: longer sims, added complexity, more difficulty with art direction and so on. I suffer from sentimentality and find the tears of FX artists unsettling, so I wanted to reduce the pain.
Can you give us an example?
We have shots where the camera is pulling back ahead of Zhong Kui while the Demon King, in hot pursuit, trips and crashes through a house, destroying it. But you never really see that house being completely destroyed, or see into the interior, because we had the camera whip around to follow Zhong Kui. But the impact is still there and the story is still told. When you don’t have a lot of time, you have to get creative about how you serve the story. In a nutshell, that was the whole challenge of this show.
The characters have a distinctive look, with inner fire glowing through cracks in their skins. How was that developed?
The fiery glow was pretty much all done in comp, using ID and UV passes to augment CG passes of procedurally generated lava textures. I think it was Falk Hofmann, our comp lead in Stuttgart, who worked the look-dev of the characters underwater. We had a lot of problems with the textures we received from Macrograph when we tried to use them in our scenes, but Falk did a fantastic job of turning a rough output into a wonderful polished look.
Tim Jones, our division VFX supervisor in L.A., was responsible for Zhong Kui’s shaders and he also did a great job – so much so that Peter Pau asked us to pass our adjusted shaders back to Macrograph to be incorporated into other shots.
Beyond that, we didn’t have a lot of control over the characters. We were trying to both match and enhance Macrograph’s look, without making it so different the audience would notice. We only had a few stills of theirs to work with – I don’t think I ever saw a complete shot of theirs until I watched the final film. So we were working a little blind.
How did you approach the character animation?
Animation was extremely challenging. We were racing to get things finished so that all the character interaction FX could be done. I actually feel that in the end we didn’t quite get to where I wanted. Some shots are beautiful, but in others the weight of these giant warriors just doesn’t come across. The previs went well, but it probably took too long, and the refining of the animation just needed more time. But the circumstances were such that we didn’t have a lot of options.
For example, it would have been great to have done animation dev on things like the snakes, but we only got rigged snakes seven weeks out from delivery, and four weeks out they were re-supplied at half the length. Peter was great about this, and understanding of what we were trying to do in the time frame – for which I’m grateful – but it does make it hard to do work when you have no time to explore.
Did the deep-sea setting make it difficult to pace the action?
I would have loved to play more with the whole underwater feel, and to have integrated it better into the animation and storytelling. Sometimes we nailed the floaty feeling, other times the size and strength of the characters against the buoyancy of the water just doesn’t communicate well. If we’d had another month, I think we would have been able to make it more special.
Tell us more about the underwater environment. Did you build it all in 3D?
Yes, the underwater environment was fully 3D. One of the real challenges the Stuttgart team faced was staying flexible while being in such a rush. 3D made it easier – we could lower a few bricks here for better contact with animation, or add bits and pieces there for better shot composition, all on the fly as cameras changed.
How about the lighting? Did the underwater setting complicate things?
Above water, our lighting was heavily simplified. Underwater, the lighting interacts a lot more with the environments. While the dust passes are faked, the non-volumetric passes are lit from the characters in reasonably complete scene files.
Compositing was pretty traditional from a technical point of view, but I think the art direction and look development from Thilo Ewers, our division VFX supervisor in Stuttgart, makes the underwater sequence very visually striking. The colour and tone are immersive, and they draw your eye to the key action, even when the scenes get murky. We went through several iterations, adjusting the amount of depth, and shots were individually tuned from templated look-dev.
You talked about simplifying the demands on the FX department. But there’s still a lot going on in those shots – fire, underwater debris and so on.
I’m really proud of what the Stuttgart FX team accomplished within the time. From final animation to the time the shots were delivered, they had about 4-5 weeks – and the last animation publishes came to them with less than two weeks to spare. Patrick Schuler, our division FX lead in Stuttgart, and his team did an amazing job.
What specific effects did they add to the shots?
For every shot in the two-minute-long underwater sequence, there’s a pass of bubbles for each character’s lava skin, bubbles from nose and mouth, interactive plankton in the water, a dust/suspension FX pass and, if there’s contact with any surface, another dust pass driven via contact.
On top of that, there’s fracturing destruction with added dust trails, fine particle passes, and large particle passes. Debris that hits the ground sinks into the ground mesh and emits further dust. There’s also brick rigid bodies, brick fracturing and ground fracturing effects. Oh, and there’s blood and a complex series of water entry passes for the first water shot.
Was it hard to get all those different simulations looking as if they were happening within the same body of water?
Well, one interesting problem was trying to consolidate the fields and underwater turbulence across multiple solvers. We accomplished this by being careful about the order with which we used the solvers, and by having emissions driven by the previous solve. For example, fracturing came first, then dust, then plankton. When one sim drives another, you get a stronger feeling that they are suspended in the same water body.
Apart from this, the FX approach was reasonably traditional: low res collision meshes were cached out, along with the higher resolution ones, from Maya. We then used Houdini to handle things like plankton, a lot of the fracturing and general destruction. 3DSMax, with Fume and thinkingParticles, was used extensively for the dust, blood, and other destruction effects. Where possible, Houdini meshes were bought back into Max for lighting, but some things – like the plankton – were rendered directly in Mantra. Cloth was done with nCloth, while Hair was handled by Ornatrix attached to a proxy head in Max. We were happy that Zhong Kui has a very manly stiff beard – it made the sims a lot easier!
How do you feel about the project, looking back?
Ah, my feelings are very complicated. It’s a Chinese fantasy film, and probably none of my family will ever see it. Even if they did, they would compare it with other VFX-heavy superhero tentpoles, and from that point of view it’s hard to see the strengths of the film. And it would be incorrect to say what we achieved is as good as the work we did on Star Trek: Into Darkness, or what we’ll end up doing on the Fantastic Four reboot.
But, for 14 weeks of work, designing and building a full CG sequence that on paper was incredibly challenging – and for a Chinese market budget to boot – well, it’s pretty cool seeing it come together, having happy clients, paid artists and some genuinely cool shots for the reel.
On that note, I’m not sure how anyone ever really copes with projects that require justification in order for people to appreciate them fully. No supervisor wants to say “it was good considering the restrictions”, despite that often being the case.
But that’s a challenge that you continually face, working in China right now. We don’t have the budgets. We don’t have the same amount of time. But the ambition is certainly there and, given time, we’ll catch up.
Zhong Kui: Snow Girl and the Dark Crystal is released in the US on Blu-ray and DVD today, 4 August 2015.
Don’t miss next week’s accompanying article, where we’ll be talking to Wil Manning about the state of the art in Chinese visual effects, and discussing the challenges and benefits of working in one of the fastest-growing motion picture markets in the world.
- Wil Manning at IMDb
- Peter Pau at IMDb
- Autodesk Maya
- Autodesk 3DS Max
- Side Effects Houdini
- Fume FX
Special thanks to Joni Jacobson, Christian Hermann and Sirena Ung.