In 2023, Doi Atsushi walked us through the VFX journey of Alice in Borderland Season 2. Following his work on The Parades, he returns to discuss the highly anticipated third season and the fresh challenges it brings.
You’ve been with Alice in Borderland since the very beginning. How would you describe the evolution of the series’ visual effects from the first to the third season?
Season 1 primarily focuses on the CG extension of Shibuya Scramble Crossing and the concept of erasing people and electricity from the city.
Season 2 strives for greater fidelity to the original work, depicting Tokyo’s streets becoming overgrown and decaying while also expanding the scale of the game stages.
Season 3 sees the game’s scale expand even further, fully reflecting the worldview the director originally envisioned for Season 1.
This can be considered the culmination of the world of Alice in Borderland.
You’ve collaborated closely with director Shinsuke Sato across all three seasons. How has that partnership evolved, and how does your shared visual language influence the design and execution of the new games?
In Season 3, the most visually interesting yet challenging sequence was the giant vortex in the final stage. The challenge in Season 3 was visualizing how characters get swept into the vortex from Shibuya Scramble Crossing, the scale of the vortex formed there, its sense of speed and depth, and the vortex’s overall structure.
First, we created an imageboard for the vortex formation point. From there, we developed detailed imageboards for key shots and shared the visuals with the director.
Working with the director, I could generally visualize what he wanted quite easily. However, the process of refining the details to achieve that vision involved multiple rounds of back-and-forth.
This dynamic hasn’t changed since Season 1, but this time, using AI to create the concept art allowed us to quickly materialize the fast-paced visuals.
When creating the original game “Future Sugoroku” that didn’t exist in the original work, one major challenge before filming was determining what kind of expression would look AI-generated when creating the AI-generated footage of the player’s future. I believe we successfully achieved that AI-like look by masterfully using compositing to make it appear generated.
How did you organize the work with your VFX Producer?
Alice in Borderland was easy to break down into distinct game segments, so the VFX producer was a huge help. They identified which vendors were best suited for each game scenario, handled the assignments, and meticulously managed the overall budget.
Season 3 introduces new games that feel even more spectacular and intense. Can you tell us about the creative approach to designing these sequences while keeping the visual identity of the series consistent?
In all of Director Sato’s works, action sequences are planned without using CG. Instead, video storyboards are created using actual stunt performers to determine the shot composition.
Based on these video storyboards, all staff members hold meetings to meticulously develop the shooting plan. During these meetings, issues, new ideas, and new ways of expression sometimes emerge.
Beyond that, for elements that cannot be clearly visualized without CG, they create CG previsualization. Additionally, to share the overall vision, they create image boards for key shots.
This process was consistently followed from Season 1 through Season 3. Season 3, in particular, featured many grand sequences, requiring extended meetings to plan them thoroughly.
The game involving flaming arrows is particularly striking. How did you approach the design and simulation of these arrows, especially the moment when millions of them rain down at the end of the game?
The sequence where fire arrows fly in:
Starting with 2 or 3 arrows, then 50, then 400, the number gradually increases, culminating in an impossible number—100 million arrows—flying in. During the pre-viz stage, we tested how the arrows would look at 500, 20,000, and 100,000. We feared that exceeding 100,000 would just result in the sky being completely covered. However, by adding depth, height, and time delays, I believe we managed to avoid that impression.
We considered the unseen backstory behind the scenario: where and how would one launch 100 million arrows?
The visuals take considerable time to show the 100 million fire arrows reaching the shrine. This implies they’re launched from a very distant location, leading us to conclude they must be fired from a launch system akin to a Patriot missile.
Furthermore, to make the fire arrows appear as if they were relentlessly attacking like a tsunami, we created multiple layers of fire arrows (varying in height and timing). This made them appear as a mass of flames flying toward the target. Creating this visual effect – where they are fire arrows yet don’t look like fire arrows – was an extremely challenging expression.
Were there specific tools or simulation techniques used to manage such a large-scale FX sequence while maintaining realism and readability for the viewer?
For the fire arrow sequence, since many shots were in the same confined space, our vendor developed a proprietary automated compositing tool. This tool automatically builds a 50-60% quality composite once matchmoving and rotoscoping are complete.
While initial development took time, I believe it proved to be a highly effective tool. It enables a small team to handle multiple shots efficiently.
For the Tokyo Tower sequence, only a portion of the set built for filming was usable. To integrate that set into the actual Tokyo Tower, meticulous asset matching was required to ensure the steel truss structure looked seamless. Since the game players’ positions gradually ascend upward, this demanding, painstaking work was essential.
For the vortex sequence and the collapse sequence at Shibuya Scramble Crossing, we initially planned to use multi-shots. However, the scene data became too heavy, ultimately preventing their implementation. This remains a challenge for future projects.
Another memorable moment is the high-speed metro crash. Can you walk us through how this sequence was conceptualized and executed — from previs to final composite?
The final shot of the subway sequence was a challenging one: the train colliding with the buffer stop, overturning, and coming to a halt.
First, we explored in pre-viz what train behavior could be created without compromising realism.
We designed the movement where the lead car hits the buffer stop, the following cars push the lead car in and derail, the rear of the lead car lifts up, and it overturns while destroying the ceiling, walls, and pillars, then comes crashing toward the camera.
For the background plate, we changed the wall on the right side of the screen to pillars to add depth. While drawing the derailed cars, we also depicted the destruction of the pillars, conveying the weight of the vehicles and the sheer intensity of the destruction.
In the final composite, we added debris from the destroyed pillars, concrete smoke, sparks from severed wires, and sparks from metal friction to create a powerful, impactful visual.
How did you balance practical effects, set extensions, and full CG shots in the metro sequence to create a sense of physical danger and immersion?
For the subway sequence, we built a set featuring one full train car and half of another car for filming.
Director Sato’s GANTZ (2011) also had a subway sequence. Drawing on that filming experience, we shot without compositing until the train doors opened. After the doors opened, we extended the view outside the doors using VFX.
For this subway game sequence, the trains run parallel—a unique setting. We determined the precise distance between the two cars, balancing safety and danger, while exploring the act of jumping between them. To heighten the risk of the jump, we added the visual element of pillars passing between the cars, making it appear difficult to jump across easily.
The new season features fresh arenas and expanded views of Tokyo. What were the main challenges in updating or reinventing the digital version of the city for Season 3?
Shibuya Scramble Crossing has undergone a major update.
In Season 2, the Shibuya area collapsed and ended due to a meteorite impact.
Season 3 begins three years later, so we needed to depict Shibuya as a city undergoing redevelopment. However, a completely redeveloped city would lose its government offices, so we had to create a design that retained the impression of the original Shibuya Scramble Crossing.
Did you rely on new scanning or drone-based data of Tokyo, or were existing assets from the previous seasons significantly upgraded for this season’s visual scope?
The scans for the new city are primarily for matchmoving purposes. For areas like Shibuya under redevelopment and the collapsed buildings in the final vortex, we’ve taken numerous reference photos and texture shots directly from the actual construction sites to create the assets.
Therefore, we’re building nearly all assets from scratch for Season 2, using very few assets from previous seasons.
The Joker’s zeppelin is one of the most intriguing new elements. What was the creative process behind its design and integration into the world of Alice in Borderland?
We wanted the Joker’s airship to be larger than the playing card airships from Season 2, so we made the Joker’s canopy 16 times the area of the playing card canopies.
Consequently, making the airship enormous would have made the scale feel unchanged. Therefore, we kept the airships from Season 2 but increased their number to six to achieve that sense of scale. This approach allowed us to convey the Joker’s immense presence.
The final sequence, with Tokyo being destroyed by massive floods and a rescue along a gigantic whirlpool, looks spectacular. What were the key technical and artistic challenges in bringing this moment to life?
The shot of Shibuya Scramble Crossing rising up and collapsing was designed to create the impression of the elevated city itself attacking the protagonists. To achieve this, we depicted an impossibly high elevation—far beyond what actual ground uplift could produce—and then triggered a mudslide and flood, making it appear as if the city were advancing toward them.
The vortex sequence was a challenging visual effect because it wasn’t just simple water flow—it required a massive vortex with a 400-meter diameter, whose center needed to swallow everything like a black hole.
We had to blend the movement of the pool water used during filming with CG water motion, mixing them to create a single water flow that felt like a terrifying, colossal surge.
Therefore, for wide shots and extreme wide shots, we simulated different water movements around the characters and the water seen in the distance. Furthermore, to emphasize the vortex’s immense scale, we made the water speed faster near the camera and slightly slower in the distance.
The plate footage featured actors performing in fixed positions, but VFX required depicting them gradually being drawn toward the vortex’s center. Pre-visualization calculations determined a vortex speed of approximately 60 km/h was needed for this sequence to work. Accordingly, we assigned the plate actors’ positions within the actual vortex and constructed the spatial relationships for each shot.
However, if everything moved at 60 km/h, the motion within each shot wouldn’t work, so we adjusted the vortex speed for each individual shot.
When you look back at the scope and ambition of Season 3, which sequence do you personally feel pushed your team’s abilities and technology the furthest — and why?
The Shibuya Scramble Crossing sequence was created in entirely different environments: Season 1 used V-Ray, Season 2 used Unreal Engine, and Season 3 used Houdini Solaris. For Season 3, we adopted a USD workflow to improve collaboration between the Effects team, Asset team, and Lighting team.
I believe this approach will be effective for all future projects as well. It became a key challenge point each season.
The final vortex sequence represented a significant technical advancement for Digital Frontier Inc., as we had never previously created water effects of this scale, leading to substantial skill improvement.
How long have you worked on this show?
It’s been about seven years since the pre-production phase of Alice in Borderland Season 1.
What’s the VFX shots count?
VFX shot is about 1600 shots (include clean up plate)
What is your next project?
The next project, directed by Sato, is already in production.
This show is a sci-fi series and is shaping up to be quite an epic production.
A big thanks for your time.
// VFX BREAKDOWNS
// TRAILERS
WANT TO KNOW MORE?
Digital Frontier: Dedicated page about Alice in Borderland – Season 3 on Digital Frontier website.
© Vincent Frei – The Art of VFX – 2025
