How Milk VFX Built the Lost Worlds and Creatures of 'Surviving Earth'

Combining natural history documentary with visual effects to recreate the deep past has become a genre of its own. Walking with Dinosaurs pushed the boundaries back in 1999 and received a serious technological upgrade in 2025, and Prehistoric Planet, Life on Our Planet, Planet Dinosaur and Earth have all contributed since.

The latest is Surviving Earth, which premiered on NBC on June 11 and is now streaming on Peacock. Tim Haines, the producer behind Walking with Dinosaurs, created the eight-part miniseries for Loud Minds and Universal Television Alternative Studio, with Josh Goodman narrating and a score by Nitin Sawhney. It tracks life through six mass extinction events, from the early seafloor to the first caveman. Haines partnered with Milk VFX, Lola Post, FIN VFX and Floating Rock to bring to life the massive environmental extinction and rebirth events that have reshaped the planet, with Milk delivering roughly 1,000 visual effects shots across the eight episodes.

"Surviving Earth is not just about dinosaurs," notes He Sun, Visual Effects Supervisor, Milk VFX. "It's a lot more variety than that. It features six extinction events and goes from the early seafloor and all the way to the first caveman. There are a total of 73 types of creatures and 55 variations on top of that, including insects, mammals, reptiles, marine life and dinosaurs."

The production adopted a more naturalistic approach to creature behavior. "This is a lot more locked-off cameras and studying the creatures rather than directing; that was quite important," remarks Joe Tarrant, Animation Supervisor, Milk VFX. "There's a narrative structure to the story, but as animators, what you're trying to do is to catch things in their most natural way in their most natural environments. Whereas in Jurassic Park it's got to burst through that wall and the cameras are dynamic. Surviving Earth is not like that at all."

Not every shot had previs. "Previs was done only for the key sequences because one of the challenges of previs is you've got to know the location first, and have to have a recce [scouting] team and get the data back in," states Sun. "Then you can have an accurate previs by knowing where you're going to actually be filming this. The problem with this show is that a lot of the places were completely remote, and we did not have the time to send the team out there for every location to be able to do a scout, come back and previs it, and then go out to shoot. A lot of this was a balancing act. The key for us was to make sure the narrative hit the beats."

He continues, "It is a documentary style, but there is narrative within that, which is what makes Surviving Earth slightly different. It's like a docudrama. There is emotion. A migration is happening, but there is a family that you have to follow along rather than just different shots which combine together to make this work. When we go onto location, there's quite a lot of ad hoc stuff going on. In the wild, you can't control the environment and the lighting. You have to capture what you can within that time period. We did have a blue cutout representing what the CG creatures are going to look like in terms of the size and the volume. It also helped to frame the shot because the directors and DPs could look at something that's not there. They did the puppet pass, took the puppet away, filmed a clean plate, and Joe and his animation team animated on top."

Natural history documentaries have their own camera language. "We had a camera stack so we would have a secondary camera underneath our main camera," explains Tarrant. "That allowed us to zoom in and do our rotations on a separate camera, so the plate remains where it is. If you need that follow-through or for the camera to lag, you can actually do that in some of the shots. Another thing we did was the arena setup, where we would have multiple cameras placed in different locations, with lighting and creatures. What you have is about a minute's worth of animation in this arena, with maybe 30 cameras set up."

Tarrant adds, "Tim or the director could look at each of those cameras and go, 'I quite like it from that angle. Could you render 200 frames of that?' That was one way we got around the cameras to try to give it that documentary-type feel. The glacier fight is a good example. We had something like 25 different camera angles and the fight itself was about a minute and a half of animation, and maybe about a minute that made the show. They pick and choose what angles they want, and then we can render those."

Milk VFX has incorporated Unreal Engine into its pipeline. "This show escalated the whole Unreal Engine pipeline for us," notes Sun. "Unreal Engine was used at the beginning of the show, when the script was even in the writing stage. There was one episode, the Carboniferous Forest, which we call Coal Forest. The forest doesn't exist anymore in modern days. The species and the vegetation inside that environment - you cannot go out to a redwood forest and shoot in California. It wouldn't be accurate at all. They had to build a 12 by 12 physical set inside a hangar as a stage. Unreal Engine was used to help the production design team to visualize the entire set while or before it was constructed. The specific trees and what sort of pattern the tree bark had were done in Unreal Engine and given to the production team to either 3D print or make the texture look like what we have in CG."

Sun continues, "That was a huge part. Also, we developed a quite unique approach, which is Unreal Engine for finishing; that's not being widely used or discussed in the industry. A lot of the larger environments you will see in the show, like seafloors, Carboniferous forest and ice world, were all generated using Unreal Engine. Without that tech in place, most of this work would be quite ambitious for this type of production, because of the constraints of budget and time. It does escalate the whole thing to the next level, because you have a lot more freedom to create those full CG environments. Even some of the creatures were rendered in Unreal Engine as well. This put us into a much stronger position in terms of this type of work."

There was no generic rig that could be applied to all of the creatures. "The variations generally speaking use the same rig, unless there's huge scale change," states Tarrant. "Some of the variations are much smaller or bigger; those would have their own bespoke rig. It was certainly a challenge for rigging because every single rig that you received for each creature needed its own rig test. When you've got 73 creatures, you try and work as efficiently as you can."

One of the first USD pipelines for Milk VFX was implemented for Surviving Earth. "USD is great at managing variants, so you can have like the same creature and have a different texture variant," says Sun. "You can use USD to switch the variant of the model or texture variant. That's managed within the USD hierarchical structure."

Damage such as wounds had to be tracked for continuity. "It's clear in the Ischigualastia episode," remarks Sun. "The juvenile got hurt and had a scar on the forehead, and that was kept all the way even when he became an adult. There is definitely a continuity of things that we constantly had to make sure it was flowing between the sequences, such as not using the wrong variant in the wrong shot."

Creature design started from the need to emotionally engage the audience. "The shapes of the creature, facial expression, eye shapes and the colors on the eyes were taken into consideration to drive the emotion to the maximum level," explains Sun. "If we know the episode starts with a happy family with newborns, the colors, tones and animation seem to be livelier. The music also helps to drive the emotion. Then you have the asteroid hit the Earth and suddenly everything goes dark. You have the effects on top, plus the emotional animation that Joe put together. Every department has to escalate the picture rather than just one part of it. The key is to get every little element combined together to make the shot work."

With no lines of dialogue, emotion is conveyed entirely through physicality. "Tim would call them the Pixar moments," reveals Tarrant. "We weren't to push it as far as Pixar, and we weren't to try to anthropomorphize these creatures, but you need those moments where you're on the edge sometimes of scientific consensus. You try to extract as much as you can from the performance, especially when the story needs it. We have to be respectful of the paleontology community, because they're very passionate. You have to make a story that excites tomorrow's paleontologists, and our job as animators is to try and be as factual and correct as you can, but also try to tell a story that does dramatize this documentary. Sometimes we do push right to the edge of things, because the story requires it, but we try to be tasteful."

Throughout, Haines pushed for collaboration. "A lot of the time, Tim relies on Joe to come up with something that makes it editorially work as well," notes Sun. "That's gold dust for us to have that creative freedom for Joe and his team, and to have a bit of fun. We offer something up, and Tim is like, 'That's great. Maybe you can change a little bit here and there.' Joe goes back, and there's another pass. We landed it pretty well. Editorial gives us a duration for the shots, but sometimes we found it is better to be longer or shorter. They are quite open for us to be able to drive the editorial with them."

Milk had to construct 100 environments. "We shot in a redwood forest, but there was a specific plant that it did not have," recalls Sun. "We had to augment that on top of the plate material. That's one type of environment work. The other one is the gigantic Prototaxites, which are these prehistorical huge mushrooms that grow seven meters tall. The plate photography was shot in Iceland. We built the Prototaxites mushrooms in Unreal Engine as well as inside Houdini USD. It was a hybrid approach. The second type is a full CG environment, like the Carboniferous Forest. The establishing shot was completely created inside of Unreal Engine. There was no plate material. We had to hybrid quite a lot to manage between the assets and the environments scene by scene and to be truthful for the scientific community as well."

The team worked practical atmospherics into the plate photography. "When the asteroid hit Earth and the dust is coming toward the camera from the valley, that was partially done through physical effects," reveals Sun. "We shot an element against black and composited it in. That helped quite a lot. The second one was large simulations like tsunamis and when water is going over the mountains and moving the earth, rock and trees. Because we had plate material, the scanning of the actual location becomes quite crucial as well. We did have a dedicated scanning team so we could replicate the environment inside 3D software to be able to carry on those large simulations as accurately as possible."

One complex sequence features stampeding Edmontosaurus and Nanuqsaurus. "Scientifically it was suggested the Edmontosaurus runs at 30 miles per hour, so they set up the camera on a car that was traveling at 30 miles per hour," states Tarrant. "There were a few ideas about how an Edmontosaurus runs, like on two legs or hops, but the one that stuck with the camera perfectly at 30 miles per hour was the gallop; that's on the edge of the scientific consensus but it felt the most natural. As we are having fun with these scenes, we are working in a scientific way as well to try to make sure we are within the lines of what could have possibly happened. A lot of fun."

Trevor Hogg is a freelance video editor and writer best known for composing in-depth filmmaker and movie profiles for VFX Voice, Animation Magazine, and British Cinematographer.