For “Gemini Man,” director Ang Lee’s latest 3D high-frame rate experiment (120 frames-per-second at 4K resolution), Weta Digital needed to elevate its animation prowess with the industry’s most believable digital human. It’s 50-year-old Will Smith (Henry) fighting 23-year-old Will Smith (Junior): aging assassin vs. his younger clone assassin. “We didn’t even think of it truly as de-aging; we thought of it as a complete creature replacement,” said Guy Williams, Weta VFX supervisor.
Junior, then, represents the logical evolution beyond Weta’s remarkable Caesar (Andy Serkis), while also applying lessons learned from “Alita: Battle Angel” (with Rosa Salazar as the titular, doll-like cyborg). This included leveraging the much lighter footprint for onset facial capture along with faster and more interactive models, better facial rigs, and greater real-time tools for instant feedback, among other advancements.
But it also required radical innovations to meet the scrutiny of high-frame rate: “Part of the challenge was the youthification of Will Smith,” added Williams. “That was exceptionally hard because Will Smith has aged so well. There are not a lot of anchor lines to scream at you what is young about him. So we had to do a really deep dive into understanding what youth really means.”
Popular on IndieWire
Weta began by making a current Will Smith CG model and then bent that down for the younger version. This was an important foundation for the animation. “We expected things to change, like the cheeks don’t sag as much, creating jowls,” Williams added. “But the nose and the ears are supposed to get bigger with age, and they didn’t really change that much on Will. He didn’t have that many wrinkles that we had to iron out. By the time we did our first four or five passes and got to the point where we were happy with the Junior model, we then compared it back to the old model, and, with no textures on it, you were hard pressed to tell the difference.”
Constants, such as bone structure, became evident along with subtle changes to the chin, jaw, and cheek. And yet the lips actually got smaller with age. And Weta noticed that Smith had no age spots to remove. In terms of the body, it’s not quite the same trapezoidal shape, as it gets older. A few modifications were made and the animators removed wrinkles and blemishes.
But Weta needed to make great strides in both skin and eye animation for Junior to succeed. “The first bridge that had to be crossed were the eyes,” Williams said. “As good as they’ve been done, even with the latest technical advancements, they still tend to be doll-like. We just deep-doved like crazy on the concept of the eyes.”
Weta did photo shoots with macro lenses filling an 8K frame to view all of the available visual information, and one of the target areas explored was how the bottom eyelid meets the eye. If you get a sharply defined, delineation between the skin and the sclera, for example, then it starts to have that dreaded dead-eye look. “Our shaders were doing the right thing, but why does a real eye look like it has a softer transition in that area?” Williams asked.
“We discovered all sorts of things that we rolled into our models and shading that really helped,” he continued. “One was the shape of the eyelid is incredibly important. It doesn’t roll down into the eye, it rolls up and then down into the eye. And the tissue gets thinner along the back of the eyelid so that it gets translucent, and that helps with the merging. And we realized there’s [a lining] called the conjunctiva, a very thin, very translucent membrane, which is like a jellyfish that sits on top of the eye. This causes veins to float within the conjunctiva. We used this info to create a slight yellowing toward the corner of the eye.” This kind of precision made the eyes much more realistic-looking.
The other critical breakthrough was in the area of skin pore simulation. Weta couldn’t just paint texture maps into a subsurface model and call it a day. A lot more sophistication was needed to pick up the true shape of tiny pores. Therefore, one of Weta’s shader writers came up with a program for simulating the skin as pore sites. It poked little holes and then used connecting cables to create the pores. Then, they used flow fields to help the pores move along the face. “By simulating the pores and growing them properly, we finally had the true curved shape of the pore so that it never flattened out at any point,” Williams said. Through the use of a mesh, this is how they created wrinkles.
The last and most important part was Junior’s animated performance (Smith played him on set and in the motion-capture volume). This involved both action and emotion. Lee wanted more visceral fighting with the stunt crew in keeping with the high-frame rate experience. He called this “Messy Fighting,” with the CG versions of Smith inserted into the action. The best example was an underground catacomb fight between the two assassins. “We did mocap of the entire fight, we matched everything up, and then we went in there and started changing the timing of the punches so that it didn’t feel like there’s a rhythm to the fight,” said Williams.
“We changed the length of the punch so that it actually passes through the guy’s face. We changed the way the guy moves his head, so that he doesn’t start to anticipate the punch moving in, and he takes the punch full on and then his head moves. Ang also wanted it to feel like a real fight where every punch doesn’t land perfectly with a lot of grazing across the chin or the throat.”
Emotionally, Smith didn’t hold back as Junior, giving a nuanced performance, such as confronting his father (Clive Owen) about his actual identity as a clone. “It fell upon us to live up to that,” Williams said. “With resolve and making sure that every pixel looks like it’s in the right place and then, when you step back, you feel the same gravity of emotion that Will was putting into the performance. If you don’t, you go back in and tear it up and figure it out. That takes a lot of time and a lot of talented animators.”