Skip to main content

How Avatar: The Way of Water's VFX Were Made

Dan Barrett takes us through his process of animating the Na'vi and creatures in 'Avatar: The Way of Water.' From capturing human emotion through Facial Performance Captures (FACS) to the complexities of working with water, the Avatar animator discusses his work with James Cameron in the newest installment of the film franchise.

Director: Funmi Sunmonu
Director of Photography: AJ Young
Editor(s): Sammy Cortino
Talent: Dan Barrett
Producer: Ashley Hall
Line Producer: Jen Santos
Associate Producer: Emebeit Beyene
Production Manager: Andressa Pelachi
Production Coordinator: Peter Brunette, Kevin Balash
Director of Talent: Lauren Mendoza
Camera Operator(s): Lucas Vilicich
Sound: Paul Cornett
Production Assistant(s): John Brodsky, Gee Depratt
Art Department: Leah Waters
Post Production Supervisor: Edward Taylor
Post Production Coordinator: Andrea Farr
Supervising Editor: Kameron Key VF
Assistant Editor: Billy Ward

Released on 03/03/2023


Facial animation for an animator, even with our tool set

would be a good three to four weeks.

If we don't get that right, then simulations break.

[intense music]

I'm Dan Barrett, and this is how we created

The Na'vi and the creatures in Avatar: The Way of Water.

[intense music]

Any animator before they start a shot,

they give us the ground and we're gonna animate

on that ground and you're not gonna change it.

But obviously with water being liquid,

we start with something that we think is gonna be, you know,

the correct surface, the thing

that we can reliably animate to

but things can change very quickly,

or it's really critical

that any interactions with a character

or a boat or a creature with the water,

it's really critical that we get the buoyancy right,

that we are inferring mass correctly.

We did have tools that would simulate

to the water, but yeah,

it's a really tricky thing.

It really did present challenges, especially that interface.

I think underwater, not so bad.

Obviously above water, you're okay,

but as soon as you get a creature or a character

perhaps even falling into the water,

there's a lot of work that needs to be done.

A lot of simulation, a lot of back

and forward between the effects team and the animation team.

Water is, I think, 800 times denser than air.

So the way that something travels

through the air versus what happens

to the travel of that object when it hits the water,

it's something that I think viewers are used to seeing,

If we don't get that right, then simulations break.

So, that's a real challenge.

You can also see that there are, you know, actors

are sitting on bucks and we are capturing their performances

and there are people

moving those bucks to try to infer the motion

that ultimately is gonna be seen in the shot.

But, really when it comes

down to it and the shots being created,

the creatures are gonna kind of do what

Jim may want 'em to do later.

So, one of the challenges is taking those performances

and then making them work on

a different bit of motion.

We've captured it on a buck that's doing this

and Jim may want it to do this.

He'd still want that performance.

So, the animators have to very carefully

use their skills to take that performance,

preserve that performance

but make sure that the body

of that character makes sense physically on the creature.

There's a bunch of work to be done there for the team.

[Actor] Woo. Woo.

[intense music]

Fun shot for an animator to work on.

And the Payakan's performance is based

on reference that we had of Humpback Wales breaching.

There was some performance capture here, but a lot

of this is the animator's work on what Low Act does.

What you're seeing in the breakdown is you're

seeing the various simulation engines.

Water's a really complicated thing.

There are sort

of multiple states that we identify.

The large mass of water when it's disturbed by a whale

and then you get a sort

of a smaller spray and then through to the little droplets.

So, what you can see here is it broken

up into those component parts.

These simulations, they take an awfully long time

and having them broken up,

it's slightly easy to have control over

the way that

the simulations would look.

The most challenging thing to animate in Avatar,

the thing that we consider to have

the most importance are the facial performances.

This scene here was one of the first that we worked on

but this was the first time we really tested

out the new system.

The face cameras that are sitting

on the booms, on their helmets.

There are two cameras, whereas

in the past we had a single camera.

The two cameras gave us the ability

to recreate the actor's faces in a 3D environment.

So that's what those cameras gave us the ability to do,

to reconstruct the faces.

And we used to use a thing called blend shapes

which was based on the surface of the face,

and we'd look at the face

and we'd infer what muscles were being fired.

But what we would do is create these shapes

in the face and we cared more about

the surface rather than the interior,

the muscles of the face.

In the new system, we had

a database of the expressions that these actors

were capable of or were most likely to use during the film.

We gathered those expressions in sessions that Jim had

with the actors where he sat them down

in a booth where they were facing an array of cameras

and he went through their entire role

in the film with them as they sat there.

Amazing to see someone like Zoe go through the emotional arc

of her performance in the film

all within a two hour session

in a chair.

That gave us all of the expressions

that she was most likely to use in the film.

And the new facial system took those expressions

and used them as a database

that it would search when it received the performance

from stage, which is what we are seeing here.

You can see these performances

and you can see how subtle they are.

There's so much range and so many of the performances

in this film, the new system did a very, very good job

of capturing all of that nuance

and giving the animators a fantastic starting point.

Animators would then look at the reference cameras

that we're seeing here.

In the reference cameras,

you can really read emotion much more easily

than you can in a face cam.

So the animators would use

these reference cameras to get a sense of that emotion

and make sure that we are getting all of that subtlety.

And a lot of that is in the body as well.

With facial animation, it's as much about the head

as it is about the lips or the eyes.

I remember it took us a long time to get this shot of Jake

where it really just wasn't reading that well.

It was lacking so much in of the intensity

that Sam had in his performance.

And it took us a few goes looking at it

before we realized that his head was just a little off axis.

It was a couple of degrees off at most,

but it broke the entire shot

and it took us a wee while to figure that one out.

[Speaker] Avatars only go around [character squeaks]

[Avatar #1] Whoa,

[Avatar #2] Whoa, bro.

[Avatar #3] Uh huh.

[Avatar #4] Go monkey boy.

The performance capture suits that they wear,

they have dots on them and these dots can

either be reflectorized or they can be active,

so sometimes you'll use an active LED.

In this case, these are infrared

and these are passive reflectors that capture

all of the body data.

In that virtual space,

what you can see is, we'll also scan all the actors.

So, you know, we'll have a Sigourney scan.

We'll have a model of Sigourney

that has her exact proportions.

Those dots you see in that virtual space there,

they sit in the same world space

as you see those dots on the suit that Sigourney is wearing.

So, that's something that we use during

the solving process just to make sure the data has come

through cleanly and we'll take that then from there

and we'll push it through onto the characters.

10 seconds of facial animation for an animator,

even with our tool set would be a good

three to four weeks of careful animation.

And then beyond that,

there's everything else that needs to go into that shot.

There's simulation for clothes and hair,

the environments, the lighting, the renderings.

A shot or a sequence of that length,

you would expect to see that going through the pipeline

for a good six to eight months.

The system that we've developed

at Weta is called the APFS.

That is an acronym for anatomically plausible facial system.

Ropes. Ropes are hard and tails are hard.

We had to animate about 4,000 tails,

at least 4,000 tails on on Avatar: The Way of Water.

All of the Navi have tails.

The Kaya people, the forest people with

the more standard cattail

and then the Metkayina with their crocodile tails.

And so, the tales are always a particular challenge

for animators.

So what does her heartbeat sound like?

A lot of the success of the production schedule

was all the planning that took place

with the team at Lightstorm before shooting began.

And one of the interesting things

about the way that this film was shot is Jim kind

of shot it twice.

He shot it with

all actors in their motion capture shoots

with the performance capture side of things.

And then a lot of it he would shoot again

in live action.

In the past, you might have someone like maybe me

on set who was hopefully holding a tennis ball

in the right spot

but the number of times that fails

and you are having to fix things,

nothing worse for a compositer here

than can you just change that eye line,

can you change those eyes on the plate?

It's a really hard thing to do.

Yeah, it was a real success.

My favorite creation, I think would be,

I like Quaritch.

I think the way that Stephen Lang played Quaritch,

I think that he's such a great character

and I think we see a little bit

of nuance in his character in this film.

[Producer] Only two shots

in the entire film don't contain visual effects.

Can you share with us which two shots

in the movie don't have visual effects?

Those shots didn't come across my desk,

so I haven't seen those shots.

[producer laughs]