logo80lv
Articlesclick_arrow
Talentsclick_arrow
Events
Workshops
Aboutclick_arrow
profile_login
Log in
0
Save
Copy Link
Share

Full Circle: Building Terminal, Escape from Tarkov's Final Location

Battlestate Games, the developer of Escape from Tarkov, told us about the challenges of building the new map with low lighting and discussed the tricky process of creating character animation. Read the interview to find out more and find a code that can get you a set of useful in-game items.

How Terminal Started

There's a curious detail about Terminal's development: our art lead Maxim Kostelov, who worked on Ground Zero, the location where every new player's story in Tarkov begins, also ended up building Terminal, the game's final location. By the time Terminal shipped, it felt like a circle closing for him.

1 of 5

"It warms my soul that I made both the beginning and the end," he says. Terminal is where that story ends. You reach it only after finishing the full campaign, by transiting from Shoreline within a limited time window. Once you're through, it stops behaving like a normal raid: a port under siege, several factions converging at once, and one way out through the fighting to the Zubr, the hovercraft that finally gets you out of Tarkov.

1 of 5

The Brief

Terminal's roots go back further than most players would guess, to a rough sketch of a port at the edge of Shoreline, drawn in 2016 by concept artist Eugene Shushliamin. When it came time to build Terminal, we started from that same sketch plus a thirty-minute brief from Nikita, our Game Director.

Thirty minutes doesn't sound like much for a location this size, but Nikita had already walked the whole thing in his head, clearly enough that it felt like he'd run through Terminal more times than anyone could count before saying a word out loud. He knew roughly where the Zubr needed to sit, where the mortar fire should land, how long the run through the location should take, and where Black Division needed to enter the picture.

Some of it he sketched by hand, with arrows for where Scavs should run and approximate marks for the strike zones and the exit point. You can trace much of the finished location back to a handful of hand-drawn lines.

Terminal also connects to Escape from Tarkov: Raid, our live-action series, though probably not in the direction people assume. Raid's production leaned on what we already had in the game. The sea terminal building that anchors Terminal, for instance, started out on Shoreline years earlier, complete with watchtowers, a concrete wall, and a checkpoint, and Raid's crew copied pieces of it for their own sets.

When we came to build Terminal itself, we already had our own understanding of the location from Nikita's brief. We looked at Raid mostly to see how a few ideas, such as the Zubr and the mortar strikes, had already been brought to life in a physical, filmed space. Even then, we made deliberate changes. In the show, the Zubr sits right next to the sea terminal building, while in the game it's a long run away, an adjustment Nikita called for from the very start.

Building a Port with No Roads

With the brief in hand, we wanted three things above all: for it to always be night, for the rain to never stop, and for Terminal to feel like a working port rather than a set piece. Getting there meant leaning heavily on reference. Ports look alike wherever you go, so instead of copying one real place, we built a composite, pulling in Barcelona's port along with a handful of others worldwide.

At its peak, our reference board ran to roughly fifteen gigabytes of images, and combing through it was one of the most enjoyable parts of the whole project for some of us. Early construction followed the same logic: stretch a reference photo over a simple blockout building, and you can judge scale and mood immediately, before any final asset exists.

Then came a problem all that reference couldn't solve: ports don't have roads the way a city does. Streets are what lets you cleanly split a city map into districts, each with its own courtyard and lighting pass, a pipeline we'd sharpened on Streets of Tarkov. A port is one continuous, flat surface end to end, which made dividing Terminal into workable zones noticeably harder. We landed on roughly twenty zones in total: two aggregate yards, a container terminal split across several sub-zones, an administrative block, and a trade complex modeled in part on a real port building.

For the first couple of months, Maxim carried the work alone, running the whole map on foot with a stopwatch to test pacing. "There was no gameplay at that stage, only Nikita's brief, so I played it all out myself," he says.

After that, zones went out to the wider team and were split by experience: less experienced artists got the more contained zones, and the more experienced creatives took the most complex ones, sometimes two at a time. The team of level artists on Terminal included Maxim Kostelov, Anton Petrov, Andrew Wells, Slava Shesterov, Maksim Pavlov, Igor Kulakov, Aigul Galimova, and Natalia Volkova; huge props to them.

A fair amount of what's scattered across those zones isn't new. Some containers date back to Hired Ops, an earlier project in the same world. Those were retextured and got a second life here. They sit alongside plenty of set dressing built specifically for this port, like the cranes and loaders.

What Changed Along the Way

Not every one of those zones survived in its original form. Two container zones, one near the oil tanks and one on the opposite side of the port, ran long enough in practice that a player had to jog the full distance out and back, with the pacing sagging the whole way.

Once that showed up in playtesting, we closed the first and cut the second in the middle. That wasn't an easy call, since the first had a specific, lived-in atmosphere that some of the team liked, Maxim included. "At first I was very upset," he says, "but then I understood it was for the better. The map plays nicer this way."

Corrections like that had room to happen because Terminal, once blocked out, sat largely finished on the art side for close to a year before the rest of the pipeline caught up to it. That stretch brought closer collaboration between environment art, narrative, and animation than we'd had on any single location before, and ideas that began as a back-and-forth between teams made it into the shipped location. It also left time to keep polishing and, increasingly, to have some fun.

Terminal has an unusually high density of memes and references tucked into its geometry. They include a shipping container gag lifted from a real, widely circulated port meme and an inscription on the stern of a large ship near the exit that is a nod to an old superhero film. We never counted how many there are in there, and we'd love to see whether the community ever catalogs them all.

Making a Night Port Run at Speed

With all of that in place, there was still a lot of technical work left. Terminal ended up larger than we expected, which pushed us toward a corridor-like layout. Unlike our more open locations, it guides players toward a specific endpoint and leaves little room to roam. Some games solve that with simple environmental cues or on-screen hints. We needed the guidance to read as part of the real geography rather than a basic overlay.

That meant building far more custom, one-off content than we were used to, such as cutscenes, scripted sequences, and bespoke lighting. Much of it ran into something our existing tools had never been asked to handle, so the routine became simple: build something Terminal-specific, find the gap in our tools, and fix it.

On the tech art side, our lead Rudolf Tsimmerman handled the overall direction for the cutscenes' lighting and effects. Our VFX artist Igor Smekalov implemented the map's interactive elements along with the cutscene effects placed in the environment.

Faces were the clearest example. Our render pipeline was never built for cutscene-distance faces, so Rudolf built a custom lighting rig for the location's cutscenes. It uses as many as sixteen light directions in a single scene, some no stronger than a thin accent light, just to keep a character's eyes from reading as two dark pits on camera.

The four distinct endings pushed that problem further. Each one moves the character far enough through space that light sources and environmental effects, even snow, have to travel and reconfigure with the character, since they can no longer stay fixed to the map.

The big explosion in the endings brought a different headache. On lower-end hardware, heavy texture compression produced visible banding where there should have been a clean blast, and it took Rudolf and Igor a long stretch of experimentation to land on a format that held up.

Solving Night Without Killing the Frame Rate

Lighting ran through many of those problems, and the permanent night added one of its own. Any directional light, like the sun or the moon, costs performance. At other locations, the player chooses the time of day for a raid, but Terminal offers no such choice, so every player walks straight into full night, whether or not they came with night vision or a flashlight. The location had to stay playable for all of them regardless, without eating the frame rate.

1 of 4

Our answer was the first lighting setup we'd built for a location where it is night all the time. Instead of a moon, we placed many large, mostly shadowless lights around the main routes. None of them is too bright on its own, but together they produce the soft glow you'd notice on a city street at night. We pushed that look slightly past strict realism because the location needed to read clearly for a player in the middle of a firefight. Characters still cast visible shadows, just softer ones than usual. Rudolf set up all of the lighting on the map, and Nikita took part in the final art direction decisions on how bright it should be.

Volumetric lighting got the same scrutiny. There was a lot of it even late in the location's development, and we decided to drop most of it and retune the lighting around the overall illumination and the glow of the lens flares. Volumetrics weren't cheap, and they also brighten the image, so a player who switches them off in the settings ends up with a darker screen than one who leaves them on. "With tears in our eyes, we had to give up some of the beautiful light," Rudolf says.

Lens flares do a similar job from a different angle, making the cranes in the distance read as lit and alive instead of dead silhouettes. As Rudolf puts it, a flare is "an impostor of light, a quad on your screen that pretends to shine." We covered the whole location with them, and when we first profiled the map, they alone were taking about 40 percent of its performance, because our systems had never had to deal with that many. Rudolf found the problem and led the fix, and graphics programmer Mark Shmidt rewrote the flare system, this time with proper visibility culling. The fix now runs on every map in the game.

Cubes, Culling & One Very Big Cube

Culling turned out to be the answer to more than lens flares. Everything visible in our locations is managed by a custom culling system, broadly similar to Umbra, the occlusion culling that comes with Unity, but built and tuned in-house. It starts with a bake across the whole map that takes twenty to thirty hours of machine time.

The bake produces a dense field of reference points, which compress into a sparser grid of cubes, roughly three meters on a side in a location of Terminal's size. Each cube records exactly what should render from inside it. Step into the next cube and the system already knows what to show you, because neighboring cubes are preloaded ahead of you, at least under normal movement.

Cutscenes broke that assumption, because a camera cutting between points outruns the preload in a way normal walking never does. Early tests of the opening sequence showed clear pop-in, with geometry racing to catch up as the camera jumped five to ten meters, several cubes at a time. The answer was one enormous cube covering the whole starting area, but there was a catch: the system couldn't do that.

Our culling had never handled cubes larger than about six meters, so we had to add that capability first. Only then could we build the cube, bake it once, and stop asking the standard grid to keep pace with a camera it was never designed for. The work also left us with a convenient draft tool, since a rough bake can now use big cubes instead of many small ones.

Culling helped with bots as well. Terminal was only the second location, after Streets of Tarkov, to get full visual culling for bots, later extended to animation as well as rendering. The system had its limits at first and only paid off on a map with as many blind corners and broken sightlines as a city has. And this is the kind of layout Terminal has. The same logic now covers every map.

The open pier at the end of the run needed its own pass. After the intensity of the fighting before it, players tend to relax the moment they step onto that open ground, and the last skirmish waiting there can end their raid. Dying to a lag spike in that fight would be especially frustrating, so we shortened the sightlines on the pier and hid the far shore, which bought back performance.

Learning to Direct

Cutscenes kept making technical problems come to the surface, and they were just as new a challenge for the people directing them: our animation lead Valentine Menshikov and his team of Svyatoslav Bazhenov, Mikhail Tsigler, Alexander Cheshev, Andrey Antonov, Egor Kozmenko, and Dmitry Libko. Everyone on that team had directed a small-scale personal project, working with scenes and cameras in Blender and Source Filmmaker or behind a video camera.

Cutscenes at Terminal's scale were another matter. The location called for several complete cinematic sequences: an opening at the base, the attack, and four distinct endings, which Vlad Nagumanov storyboarded. All of it was built on existing deadlines alongside our ongoing work on the interactive traders.

Character animation itself was never the obstacle, since we already had experience with it. Facial animation and capture, rigging, and retopology were the real unknowns, and we picked them up largely on the job. Svyatoslav Bazhenov took on much of that technical work: facial rigging, capture process, some retopology, and unifying the face meshes of the cutscene characters.

"All the guys did great," Valentine says, "Everyone tried their best, regardless of the experience they started with."

With little time to train, we held ourselves to one simple standard: make it look great. For reference, we studied the cinematics of Battlefield 3 and the original Call of Duty: Modern Warfare games, Call of Duty 4 and Modern Warfare 2, for how they build spectacle into a scene.

The firefights in the 1995 Heat movie and the 1999 intro of Resident Evil 3: Nemesis were on the list too, alongside our own Raid series for tone. The start of the attack sequence, with the shelling and the Scavs pushing in, is an emotional homage to the ending moments in Raid. We were never asked to copy the series shot for shot, and in this scene we tried to recreate its feeling as closely as we could.

Chasing that feeling meant a hands-on pipeline. We animated in Maya, then assembled and finished the scenes in Unity's Timeline, with a large share of the polish, timing, and camera work done by hand. "The hardest part was editing, even more than animating," says Valentine, who put the scenes together in the engine.

Getting the scene's rhythm right meant cutting it, watching it back and refining it again, over and over, before we could call it finished. Each scene also had to be synced with programmer Dima Kuprik and sound engineer Daniil Sedelnikov. We showed Nikita and our designers small pieces along the way, never a whole sequence at once, and adjusted each piece based on their feedback before building the next.

The Detail Nobody Storyboarded

Plenty of the best moments in these cutscenes were never storyboarded. The opening sequence, the walk through the checkpoint and into the base, was framed at a basic level by Polina Vilimok, and the camera angles and movement were worked out as we went.

We wanted every background character to feel like an individual, so we invented them on the spot and motion-captured each one separately. One guy chats with another, who is talking to a third guy tinkering under a UAZ truck, and an officer taking notes glances up as a crate gets loaded behind him. We added all of it along the way because it made the space feel alive.

The base's reaction to the player was staged just as deliberately. Heads turn and follow you the moment you walk in, the way any real checkpoint would eye an unfamiliar face, and the tension drains out of the crowd once you're waved through. The idea was for every face to ask the same question: who is this dude? In our understanding, the local peacekeepers have little respect for BEARs and even less for USECs, hence the tense glances from every side.

A wider aerial pass over the attack sequence got the same treatment. It was a complicated scene. Polina did its basic framing as well, while Rinat Rafikov and Stanislav Vintage from our video production team helped develop and correct the early sketches. A small, densely choreographed story plays out in the background: Scavs scatter from a blast, and one character lies in a deliberately silly pose that nobody's sure anyone has noticed. It's the crater pose from Dragon Ball Z, an idea from animator Dmitry Libko. Now's your chance to look.

Motion capture had its own complications. We used Perception Neuron and Xsens for motion capture, and Perception Neuron is sensitive to magnetic fields. The wiring under our office floor interfered with tracking, so even a good take sometimes needed manual repairs afterward.

For faces, we had no professional capture rig at the time and improvised two: one was assembled from construction materials we had on hand, and another was built around a tactical helmet and a smartphone. We recorded facial performances in five to ten takes, picked the best one, and reworked it heavily by hand, so nothing in the cutscenes is raw capture. The traders use a lip-sync plugin instead.

Not every scene was a big one, and the finale shows it best: two or three characters in one small boat, a far quieter frame than everything before it. After the scale of the fight it takes to reach it, that restraint feels earned.

Conclusion

That quiet final beat is where Terminal ends as a location, but what we learned making it carried over into later work. Icebreaker is a good example: its cutscene work, including the reveal of Black Division, moved faster and landed with more polish because a working pipeline had already existed by then. Rinat, Stanislav, and the rest of our video production team had also supplied detailed live references and storyboards.

The payoff we felt most came from watching someone else experience the location. When Tigz became the first streamer to play Terminal, seeing it land made the whole effort feel worth it, and the reactions from players since have confirmed it.

Terminal isn't perfect. There are things we'd still tighten, effects we'd push further, and calls we'd probably make differently with more time. Maxim sees it the same way: "It holds a special place in my heart, but I understand it isn't quite perfect. Maybe that's for the best. The best is the enemy of the good."

Many more people worked on Terminal than we could name here, and we're grateful to every one of them. Thanks for reading, and thanks for playing. See you all in Tarkov!

Check out Escape from Tarkov's final location and use our code READ80LV to get a solid set of valuable, useful in-game items.

BATTLESTATE GAMES, Escape from Tarkov developer

Built for the Game & Digital Art Industry
Get Our Media Kit

Comments

0

arrow
Type your comment here
Leave Comment
Built for the Game & Digital Art Industry
Get Our Media Kit