FACE TIME WITH ALITA: BATTLE ANGEL
The mocap, modelling and movement secrets behind the face of Weta Digital’s latest completely synthetic character, Alita
Weta Digital reveal their stunning facial mocap secrets on their actionpacked manga adaptation Alita: Battle Angel
The holy grail of visual effects is the creation of a living, breathing digital human. In Alita: Battle Angel, from director Robert Rodriguez, Weta Digital stakes its claim as one of the few studios capable of pulling off such a character.
To do so, artists at the studio took performance capture by actress Rosa Salazar – who wore a fullbody capture suit and head-cam – and translated that into the CG avatar Alita, a humanoid cyborg making her way in a post-apocalyptic world.
Central to the believability of Alita was her face, an area of digital characters that Weta Digital has perfected over many characters in several productions, from Gollum in the Lord Of The Rings and Hobbit films, to the Planet Of The Apes primates. Here’s how they used the latest tools and techniques to make this newest CG character possible.
CAPTURING THE FACE
On past productions, Weta’s facial capture has generally been facilitated by a head-cam setup, featuring a single camera on a boom arm that points back at the actor’s face. But for Battle Angel, the studio implemented a new suit and a new head-cam that employed two cameras aimed at Salazar’s face, which was covered in tracking dots drawn on each morning before shooting.
“The dual camera system for the face allows us to triangulate those dot positions in 3D space,” explains Weta Digital animation supervisor Mike Cozens. “This allows you to understand the distance and depth within which the face is moving. So, as you squeeze your lips together, and your lips push forward, we can understand how the skin is moving, and how those dots are moving on the face. That was a new bit of tech that our team came up with on this show.”
So why was capturing that detail so important, especially since Weta Digital has clearly been able to produce such impressive CG characters previously? The reason, suggests Cozens, was that so much of the film lingers on Alita’s face, and there also tended to be a great deal of interaction scripted between the main character and other – sometimes live-action – ones. Plus, there was eating. A lot of eating.
“Alita eats an orange and chocolate – there’s a lot of food,” says Cozens. “This goes from a live-action character handing Alita an orange and peeling it open, to her picking it up and chewing it in a close-up. Your face is an amazing thing, and watching something like that happen is a level of complexity that we hadn’t really done before. It was really just [about] breaking the face down into all the components in order to make it work.”
FACE TRANSLATION
Salazar’s motion data from the capture process was tweaked by Weta Digital’s skilled motion editors. In terms of facial capture through to animation, the motion data was firstly applied to a digital representation of Salazar’s face, with the aim to ensure that all the nuances of the actress’ performance were ‘solved’ on a one-to-one digital representation of her, before tackling the CG Alita facial animation.
Weta Digital visual effects supervisor Eric Saindon says that building this intermediary ‘digital puppet’ version of Salazar allowed for the analysis of the finest nuanced details from the performance capture of the actress. “It’s not something you’d ever notice just looking at her, but once you start capturing all that extra detail and comparing it to the 3D model, you go, ‘Oh wait, we’re missing this little detail up here’.”
This also worked well because Alita was intended to be a direct translation of the essence of Salazar’s performance. Although Alita is a cyborg – and has some manga-style features like enlarged eyes – it was the actress that the filmmakers wanted to be represented in the CG character.
“The very first day of shooting, Rosa did tend to be a little bit more cyborg in her motion,” notes Saindon, “but Robert Rodriguez really didn’t want that at all, he didn’t want her to act cyborg. She just was supposed to be Rosa. Alita doesn’t really act like a cyborg. She looks cyborg, but her mannerisms and her emotions are really just Rosa. We didn’t change anything from that. We kept it as Rosa.”
Ultimately, the facial capture data from Salazar ran through Weta Digital’s proprietary animation solver, with fine-
“Alita looks cyborg, But her mannerisms And her emotions Are really just rosa” Eric Saindon, visual effects supervisor, Weta Digital