When Marvel’s Black Widow premiered in theaters and on the Disney+ streaming service, it not only ushered in the next phase of the Marvel Cinematic Universe, but it also welcomed many fans back to movie theaters after a long hiatus.
Black Widow follows Scarlett Johansson’s superhero secret agent, Natasha Romanoff, as she finds herself back in Russia investigating the whereabouts of some former allies from the time before she became an Avenger. The film’s explosive third act is set inside and up in the sky around and below the floating Red Room, a secret training facility for assassins that explodes around Natasha, forcing her to flee, then engage in a midair battle with the villain Taskmaster.
Visual effects studio Digital Domain was tasked with digitally creating — then destroying — the Red Room for Black Widow, extending a partnership that had the studio previously work on Captain Marvel and other Marvel films. We spoke to the studio’s digital effects supervisor on the film, Hanzhi Tang, about the process of both creating the massive structure and then blowing it to bits around the film’s characters.
We: How did your work on Black Widow compare to other Marvel projects you’ve worked on? Were there more VFX shots for your team on this one, or fewer?
Hanzhi Tang: I think it was around 320 shots. It was a big show. The shot count wasn’t as high as we’ve done in the past — we’ve done 800 shot shows before — but the difficulty of this one was pretty high. So it felt like a really big task.
The Red Room is such a big part of the film’s third act, and Digital Domain played a big role in designing it with Marvel. What went into the creation of this particular set piece?
We wanted to create something massive that had scale, but was also incredibly intricate. It was going to test the limits of what we could do, so we spent the first few months just figuring it all out and doing little tests in which we crammed as much geometry as possible into the model and made sure it worked. Being able to animate it, destroy it, light it … there was a lot of testing, but eventually we were like, “I think we’re going to be OK.”
What were some of the visual influences and touchstones for the design of the Red Room?
[It was] mostly Soviet-era architecture. It’s like an oil rig in that it’s a structure that’s exposed to the elements a lot, and there’s this brutalist style of architecture to it, as well as a communications tower, too. There’s so much glass and steel, but then it’s also very rusty and beaten up.
And then you blew it up! Does it make it easier or more difficult to work on the destruction of something you also had a hand in creating?
For a lot of effects artists, that’s a dream job: Blowing stuff up. Who doesn’t like blowing stuff up? But it’s also something we, as a company, have done a lot of and have a lot of experience doing. Going back all the way to disaster movies like Day After Tomorrow and 2012, people here get very excited when they get to destroy something.
The scenes that unfold while The Red Room is exploding and the characters are hurtling through the air amid all of the debris are spectacular. Were those scenes as complicated as they seem?
Well, [the characters] are high up in the atmosphere, so that was something new. We had to consider whether they were at airline cruise altitude or some other altitude, and what that might mean — like, how high up is this, really? We also had to figure out how long they need to fall in the skydive, because that is a really long skydive. Along those lines, we had to figure out whether the wind speeds at that level would drag the surrounding smoke out, and all kinds of other little details. Like, do they need to wear oxygen masks at that altitude? Is it that high?
Trying to combine the cloudscapes into it was also new for us, as we had clouds everywhere and this atmospheric stack they had to fall through. The actors go from sunny and exposed at the top, just at sunset — basically, the golden hour — but then it gets more overcast in the middle of their fall, and finally, they come out of the clouds at the bottom and the sky clears up. So is that more of the same golden-hour lighting during their fight on the ground? The weather, the lighting, all of it had to be thought about throughout the whole sequence.
What about all of the debris falling around them? Did that complicate things?
Definitely. We had a library of chunks of debris and also some some premade assets that created something like a curtain of stuff falling, so you could create a diorama of stuff falling in each shot. That made it possible for every camera view you’re shooting through to have pieces of the Red Room falling around them.
There was also a lot of tracking of large pieces with more recognizable shapes. We had to track those shots and make sure that they stayed somewhat consistent, and a lot of pieces were also placed into shots to look aesthetically pleasing while making sure nothing got in the way of the camera.
You worked on some earlier Marvel movies, and I suspect certain elements of your process have evolved a lot since Captain Marvel and the earlier films. What was the biggest change this time around?
It was mostly the GPU rendering. GPUs are very hard to find right now. We’ve rendered on CPUs for most of the life of this company, but with the speed advances that you can get with GPUs, that’s something we couldn’t ignore. At the same time, rendering with GPUs is so new that we couldn’t immediately risk changing the the way we do things too quickly.
Yeah, you don’t want to risk that ripple effect through the rest of your work …
Exactly. And because there are so few GPUs at the company, we had to build up a whole new render farm to do it. We started off using GPU for animation renders. Those will never be seen by the public, because it was like an internal sandbox we used to see how fast we could go. That was around the time of Avengers: Endgame and Infinity War, when we only used GPUs for internal animation renders for approvals. Then we moved on to Captain Marvel, which was the first project in which we used GPUs for the final rendering. You can see some of those shots in the movie — particularly when the Skrull ship blows up.
When they’re escaping the ship near the end of the film?
Yeah, we did that hangar scene, when they bust out of the hangar and fly down to Earth. We also did the canyon chase visual effects in Captain Marvel, and all of that was rendered with GPUs. Basically, we saw all these render times that were really awesome and did some testing to make sure the quality was as good as it would be if we had not [used GPUs]. We tried to find the weaknesses in it, because no new technology has everything covered, but it looked great. So Black Widow was the next one for us.
For the Red Room, it’s all hard surfaces, metal and glass. So render-wise, we had high confidence in being able to cope with it. [Using GPUs] really saved our bacon as far as rendering this massive asset tons of times. We iterated it every day, and if you had to wait the usual two days [with CPUs] to get a full render of a long shot to review, I think we would have made life much harder for ourselves.
Is there something in the film you worked on that people probably won’t even realize is a visual effect?
As they’re tumbling through the skydive, there are some shots of them grappling, and I don’t think you would necessarily notice we’ve done a full replacement of the characters in some shots. It’s interleaved between the actors and the more obvious stunt work. They do some very difficult moves at times that likely aren’t the actors, and there’s some stuff in the middle that we just ended up replacing because it was easier. Either the actors were difficult to rotoscope with the background or we didn’t like the hair or lighting on them — maybe hair was blowing in the wrong direction or blowing over their face, or something like that. So we just ended up replacing them.
That’s one of those avenues you go down as you get the shots back and the director’s like, “I don’t quite like that pose, can you replace this or that?” And then you ultimately end up replacing more than you initially planned.
What are some of the scenes you’re most proud of working on in the film?
There are two shots inside the Red Room which are kind of one-off destruction events. One is where Natasha and Taskmaster are facing off, and then at the same time, the entire room breaks in half. That was an extreme amount of one-off work. All the exterior Red Room stuff we developed can be reused for multiple shots, but that one scene was a lot of work just for that shot, and there’s a lot of really cool destruction in there.
Later, Natasha stumbles through a corridor when the Red Room is being destroyed and there’s a lot of debris falling, and a couple of those are all CG, where we replaced Natasha as well. Those came out really nice. I’m really proud of them.
Marvel’s Black Widow is available to watch now on the Disney+ streaming service via Premier Access.