Many thanks to ehoba, whose priceless contributions to online anime history discourse have made this article possible, and to Alia Demnati, whose exemplary work was a major inspiration.
Patlabor 2 is widely considered to be one of the best anime films ever made. For all those involved, it was an incredibly ambitious and important production, far more than the first movie which had come out 4 years prior. Director Mamoru Oshii himself stated, somewhat provocatively, that, “as far as I’m concerned, the only Patlabor I’ve made is Patlabor 2. It’s the only thing that I can proudly call my own work”. As controversial as that statement may be, it is true that Patlabor 2 is a step-up in nearly every aspect compared to its predecessor, from its budget to production pipeline to Oshii-isms. Beyond this, Patlabor 2 is generally held to be one of the most important and influential works in Japanese animation and one of the defining movies of the 1990s – not so much for its extremely coherent aesthetic and deeply thought-out writing, but rather for the unique elements it introduced in the Japanese animation pipeline: a renovated layout system, new ways of using CGI and a still-unmatched ability to challenge live-action cinema techniques. Starting from the movie’s staff and production methods, this article will try to retrace how all of these elements coalesced to produce such a cohesive whole.
Building on Patlabor 1
To understand Patlabor 2 and how it came into being, it is necessary to go back briefly on the evolution of the franchise since the previous movie, which came out in 1989. Following the success of the original OVA, the first film was soon followed by a 47 episodes TV series which aired from November 1989 to September 1990; then, a new OVA was released between November 1990 and April 1992. One year later, on August 7, 1993, it was Patlabor 2’s turn to come out in theaters. For multiple years, the Patlabor franchise had been incredibly popular and the new movie was its crowning achievement – as well as its last entry until the early 2000s.
The franchise’s growth is inseparable from that of two companies – toymaker and animation producer Bandai on one hand, and animation studio IG Tatsunoko on the other. As explained in the article on the first movie, it was largely thanks to Bandai producer Shin Unozawa that Patlabor could exist in the first place. He remained on the franchise throughout all its installments, but things were starting to change in Bandai: in March 1989, the company’s animation distribution subsidiary, AE Planning, renamed itself Bandai Visual Sales, and then Bandai Visual in 1991. The same year, Bandai Visual absorbed the conglomerate’s video release company, Emotion, therefore becoming a behemoth within the film and animation industries. Patlabor 2, produced by first-rate Bandai Visual employee Unozawa, was the company’s biggest project yet. In line with those ambitions, Patlabor 2 was a much larger production than the first movie: it had double the budget – which would make it between 200 and 300 million yen, low compared to bubble-era films but a substantial sum after said bubble’s burst.
All of the Patlabor entries between the first and second movie were produced and animated by the studio that had initially refused to turn Headgear’s early drafts into a series: Sunrise. As Mamoru Oshii turned towards live-action (with Kerberos Panzer Cops in 1991 and Talking Head in 1992) and tightened his relationship with Bandai Visual (which produced both movies), there was no reason to rely on the producer of the initial OVA and movie, studio Deen. Bandai naturally turned towards its close partner Sunrise for the sequels, and would probably have had it animate Patlabor 2 as well. However, Unozawa insisted on having Oshii in charge, and the director bargained so that it would be made on his conditions: first, Bandai Visual would have to produce Talking Head, and then leave the movie under the hands of the actual studio behind Patlabor 1, IG Tatsunoko.
Between 1989 and 1993, the small company which only had 5 employees at its creation had grown. It had spent these 4 years doing in-between and key animation outsourcing for various other studios, before handling projects all on its own such as Video Girl Ai or The Weathering Continent in 1992. By 1993, IG was split in two divisions: studio 1, led by IG co-founder Takayuki Gôto, and studio 2, led by Patlabor 1 animation director Kazuchika Kise. Kise and many of the artists from Patlabor 1 would therefore be back on the sequel.
Oshii’s demand that IG be put in charge of the movie not just as subcontractor but as the company behind the entire film was a strategic decision. The first movie had been an artistic success because IG was a young studio that Oshii could shape by introducing it to his closest collaborators and setting up his own specific production pipeline. Having IG on Patlabor 2 (as well as Talking Head just before that) confirmed that it was Oshii’s studio and that their mutually beneficial relationship would last. It would indeed, and much of IG’s fame and excellence since then can be traced back to the staff that were brought in and experiments that were tried out in terms of pipeline.
While that may be less obvious and even less conscious on Oshii’s part, settling in IG was also an aesthetic decision which definitely put him in a very specific position in the landscape of 90s Japanese animation. Indeed, at the same time as Patlabor 1, Oshii directed Gosenzosama Banbanzai, a 6-episodes OVA released to commemorate the 10th anniversary of studio Pierrot. Extremely different yet similar in some ways, Patlabor 1 and Gosenzosama indicated the two ways in which Japanese character animation would evolve in the 1990s.
Under the lead of animation director Satoru Utsunomiya, Gosenzosama paved the way for a kind of animation that, although it relied on low framerates, would endeavor to be as expressive and dynamic as possible, while never losing grasp of realism in terms of depicting weight and forces. This tendency, which evolved in what I call “flow animation”, was well adapted to the constrained environments of TV and OVA productions and naturally flourished there, carried by young TV animators such as Norio Matsumoto and Tetsuya Nishio, or slightly more experienced and experimental OVA artists like Mitsuo Iso, Osamu Tanabe and Shin’ya Ohira. Although the latter three formed a cohesive group for some time, their efforts were scattered over many small and obscure productions – an obscurity which also guaranteed their experimental and radical nature, such as in Ohira’s “The Antique Shop” or the Junkers Come Here Pilot film.
On the other hand, Patlabor 1 represented a more controlled, meticulous approach that was closer to the “slow acting” of 80s realism than the expressionist animation of Utsunomiya’s followers. It insisted on graphical detail, especially in shading, clothing and body structure, and on slow, nuanced character acting. Rather than the individualism of certain experimental animators, it relied on two specific positions, that of layout artist and animation director. As a result, it flourished on movies, where those two steps of the production had the most time to develop and produce really satisfying results. By putting a major accent on those specific moments of the production, Patlabor 2 had a major stylistic influence and set a gold standard for what “movie-quality animation” should look like.
The movie therefore brought together under animation director Kazuchika Kise many artists who were moving in that tendency, such as Masahiro Ando, Satoshi Kon or Hiroyuki Okiura. Okiura’s case is perhaps the best to show how both Patlabor movies, and especially the second one, were a shift towards that new approach of animation. His animation between 1988 and 1993 is, with some nuances, much closer to the Gosenzosama current and appears to be rather close to Mitsuo Iso, such as in Sukeban Deka or Rôjin Z. However, after Patlabor 2, his animation would become increasingly methodical and photorealistic. By the second half of the 90s, on Ghost in the Shell and especially Jin-Rô, Okiura’s animation would have taken the shape it is now known for: that of a movement so physically accurate that it appears rotoscoped.
In its animation and design philosophy, Patlabor 2 explicitly set itself in the continuity of Patlabor 1 – or, rather, it pushed to their full realization all the things that Patlabor 1 had attempted. This is especially obvious in the character designs, realized by Akemi Takada and Masami Yûki (for original characters, notably Arakawa and Tsuge). On Patlabor 1, Takada had protested against the modifications that Kise had brought to her models, as he changed many elements to make them more realistic and, in the process, less attractive. But on Patlabor 2, she seems to have been convinced by Oshii and Kise’s arguments: her character designs would be completely different and much closer to the animation director’s philosophy.
The images above, comparisons of Takada’s reference sheets from both movies, give a good measure of the changes. First, characters immediately feel older, more mature and meditative – something that Takada explained by saying that, just like the staff, they had grown between the two movies. They are also far less “anime-like”, which is to say that most of the elements that made Noa and Shinobu bishôjo characters were eliminated: noses and chins became much rounder, eyes smaller, and the features of the face were generally heavier. The repartition of all the details was also very different: the faces of the characters in Patlabor 2 generally feel far simpler, although that is compensated by the extreme care put in the hair in the settei themselves, or in the shading added by Kise during the animation process.
Shots such as this one are a good example of the extreme simplicity of the character designs, with an extremely reduced amount of lines, but also barely any expressions. With such models, the shading is of paramount importance to indicate volume and also the basic shapes of the eyes, mouth and nose.
Unlike what happened on Patlabor 1, Takada and Kise willingly collaborated, for superb results. What these results consist of exactly, however, is rather hard to describe precisely. In line with the movie’s general aesthetic, they are generally described as “realistic” – but what does that mean exactly? In a way, I’d say that their most essential feature is their minimalism – the lack of details, but also of distinct expressions, which makes them particularly well-adapted to both the mysterious, slow mood of the movie, and to modifications by the animation director. In achieving that, they also managed to strike a perfect balance between simplicity in terms of visual aspect and complexity in terms of physical features. This is perhaps how the “realism” of those designs could be understood and reformulated: not in terms of photorealism or extreme detail (as would be more the case in Ghost in the Shell), but rather as certain kinds of body structures that constrain the animation on one side (they are hard to move without losing consistency) while liberating on another, very specific one (they are simple enough to work as a template for the animation director’s own tendencies).
A new layout system?
The quality and success of Patlabor 2’s animation did not only rest on its character models. In fact, most accounts of the movie’s production and impact point to another element: its layouts. Before diving into the specific role they played, it’s necessary to provide a general overview of what layouts are in the first place, and how the movie may or may not have played a role in their development. For this part of the article, I will heavily rely on a specific book: Methods – Mamoru Oshii’s Patlabor 2 Director Notes. Published in 1994, it contains an extensive collection of layouts from the movie accompanied by detailed commentary from Oshii himself, as well as multiple interviews from the staff.
In Methods’ brief preface, Oshii provides a now famous definition of the layout, which I will quote here to begin with:
A “layout” is, in short, a blueprint for the screen (a cut); the “layout system” is a method for producing a movie (or, more precisely, a film) that uses such a blueprint as its foundation.
As a blueprint, the layout contains various kinds of information: the size of the angle of view, the perspective for coordinating the background and animation (the cels), the type of lens that should be used, the frames and scales to set the camerawork, the book [indications] for the background art staff, the brightness of the image for the color design, the information needed for the animation, and so on…
For the staff that is divided across different divisions, it serves as a standard list for the work that needs to be done, while for the director, it acts as a simulation process that allows them to examine the production before the shooting.”
As we will see, Oshii placed an importance in the layout that most other directors did not. But in spite of the many tasks that he assigns to the layout, his definition points to its most essential feature: it is the document that serves as the main instrument of communication between different parts of the animation process once the storyboard has been completed.
Layouts have more-or-less always existed over the history of Japanese animation since 1958, although their importance, amount of detail and exact use varied according to each studio and pipeline. The generally-admitted date for the first clear formulation and systematization of the “layout system” is 1974, with the production and release of Heidi, Girl of the Alps. There, director Isao Takahata took away the creation of the layout from the hands of the animators to give it to a dedicated, specialized artist: in this case, Hayao Miyazaki. The goal was to provide a strong baseline for the animators to work from, while enabling better coordination and communication between the animation, background and photography.
This initial layout system quickly took hold in Takahata and Miyazaki’s studio at the time, Nippon Animation, but had difficulty spreading all over the industry. Rather than a tool for coordination, it quickly became a way for animation directors to exert greater control over their own productions: such was the way Miyazaki himself used it on Future Boy Conan or Yoshikazu Yasuhiko on Mobile Suit Gundam. It is at this point that we can mention Oshii’s first contact with this approach to layout: it was introduced to him by former Nippon animator Toshiyasu Okada, animation director on The Wonderful Adventures of Nils, in 1980. Still, in most cases, things remained the way they had been in the 70s: animators did their own layouts and directly sent them to the background artists. This was also the case in Miyazaki’s movies.
By the second half of the 80’s, however, the position of layout in theatrical productions started to evolve. On one hand, animation directors started correcting not only the key animation, but also the layouts: for example, Hiroyuki Kitazume corrected by himself all the layouts of Char’s Counterattack in 1988, while the other animation directors were in charge of supervising the key animation. On the other hand, an increasing number of movies began including specific, specialized artists on layouts: we can cite The Wings of Honneamise in 1987, Grave of the Fireflies and Akira in 1988, Patlabor 1 in 1989 and Rôjin Z in 1991. At this point, processes were quite hybrid and still varied from production to production: on Honneamise and Akira, the layouts were mostly an extension of background design, while on Rôjin Z, the distinction between layouts and key animation seems to have been blurry. As for Patlabor 1, the animators did their own rough layouts first, which were then heavily revised by the two dedicated layout artists, corrected by the animation director, and finally used as actual blueprints. Patlabor 2 rationalized that by leaving the animators out of layout creation, which became the product of the exclusive dialogue between the layout artists and the animation director.
In Methods, many of the artists from Patlabor 2 formulate the wish that layout creation would be acknowledged as a distinct part of the animation production process and that the layout system be introduced in as many productions as possible. One could argue that this has indeed happened, but in a way that Patlabor 2’s staff could not predict and now possibly regrets. In today’s anime production pipeline, layouts are a separate task, but one that has largely replaced the function of key animation and detrimentally fragmented the pipeline. As regrettable as it is, this was in fact a natural development of the situation in the early 90s.
Indeed, Patlabor 2 had quite an impact and many artists reported having read Methods and trying to implement more detailed layouts in their own work. This took shape in two ways. On the side of the animation directors, they started to correct layouts on a more systematic basis; but the consequences were that the limit between layout and key animation became blurry, and that the number of animation directors on a given work had to increase to meet the rise in workload. On the side of the animators, more detailed layouts were expected, to the degree that they progressively transformed into rough key animation. At some point, it became impossible for key animators to handle both layouts and key animation; the two stages were then merged into the hybrid “layout/first key animation” terminology while second key animation (essentially clean-up), which had until then been an attribution of in-betweeners, was officially acknowledged and explicitly credited as an entirely separate job.
Of course, Patlabor 2 or its staff cannot be held solely responsible for the situation of the anime industry 30 years after its creation. In fact, as I have briefly shown, this was an industry-wide change that Patlabor 2 reflected, perhaps accelerated, but did not and could not have provoked. Where the movie did undoubtedly innovate, however, was in the amount of information put in the layout: as Oshii puts it in his definition, it includes “angle of view, perspective, lens, frames and scales, brightness”… Many of these elements, notably the angle, perspective and lens, were specifically added on Oshii’s demand. Indeed, for him, the layout was the tool from which he, as a director, could control the final look of the film; this partly explains the many differences between the storyboards and layouts, as the former were not envisioned as the definitive version. Oshii explained the different functions of the storyboard and layout as follows:
In a sequence that requires a dramatic composition, setting up space naturally begins with the selection of a “place” at the storyboard stage. However, whether that “place” can be used effectively depends on the layout work.”
Then, for Oshii, the layout embodied the point of view of the director not as a storyteller, but as a visual artist. It also embodied a fictional object the director wanted to insert into his animated film: a camera. Wanting to imitate the look and feel of a live-action movie, Oshii refused to consider the space where the characters evolved as a flat stage, and took into heavy consideration concepts such as distance (object-to-object and object-to-camera) and perspective. Each layout also specified what lens would be “used”, or rather, what lens effect would be imitated. The layout was the perfect tool for this, since its place in the production process made it particularly fit to coordinate distortions on both the background and cels. To properly understand this, discussing a few of the examples Oshii provides in Methods is appropriate.
To begin with, we can take a look at cut 135, from a scene at the beginning of the film when Shinobu is driving, just before being caught in a traffic jam – a kind of scene particularly complex to elaborate, since it entails showing cramped spaces (inside cars) and take into consideration the different speeds of each vehicle in order to recreate the irregular, bumpy progression of cars in a jam. But for this specific shot, what Oshii’s commentary emphasizes is the role of perspective, and more precisely of the vanishing point, to create the composition.
The viewer can barely notice it, but the background and overall perspective are slightly tweaked in order to create the impression of a particular lens being used. They are “tweaked” relative to a specific point in the frame: “around the armpit”, indicated on the layout by a small cross. The result is that, the further away they are from the vanishing point, the larger (and closer) objects will appear. Here, by putting the vanishing point very close from the camera, the layout emphasizes the presence of all the elements around Shinobu. In Oshii’s words, “it enlarges the sky at dusk that fills the window of the car”. Then, all the other elements can be set in place to support the overall impression: Shinobu’s body is positioned in order to “support the perspective of the interior of the car” and is drawn while paying specific attention to the body structure. It is there that what the staff calls a “lie” slips in with a slightly off-model but perfectly planned element: Shinobu’s neck is drawn particularly thick, once again to support the composition.
Then, we have cut 238, a shot of a plane’s cockpit. Here, the focus of the composition (and therefore the vanishing point) is the plane’s head up display (HUD) – not indicated in the layout, but which was created by the CGI staff and composited with the 2D image. Because of this, this shot is repeated multiple times. Furthermore, Oshii explains that he did not want to vary too much the perspective here, as “just lining up different angles in a half-baked way tends to reduce the impact of each individual layout”.
This explanation is interesting because it raises the issue of the angles that were not used. One of these Oshii explicitly comments on: it would be to show the pilot and the cockpit from the profile, in a composition similar to that of cut 135 discussed above. Here, the relationship between background and cel layers would be completely different, because what is shown doesn’t carry the same meaning:
At least in animation, layouts where the background is the infinite expanse of the sky should definitely be avoided. Not only do we lose the sense of altitude, which is essential for depicting the cockpit of an aircraft, but the follow (the movement of the background layer) tends to ruin the dynamism and vastness of the sky, which generally results in a weak, tasteless shot.”
Finally, let’s discuss cut 86, Noa and Asuma looking up at Alphonse in the Labor testing site. Here, the camera is positioned below eye level and, just like the characters, is looking slightly up, which, according to Oshii, “is frequently used to express the distance between the figures and to give depth to the image”. However, he also mentions that such a shot can be risky to use since “it immediately exposes the false height difference if it is not consistent with the perspective. […] If the perspective of the entire body, shoulders, chin, under the nose and eyes does not line up in accurate drawings, the viewer will feel uneasy and the composition will lose its sense of weight.”
In this case, we see how my earlier remarks on character design and the animation work in general come together in this specific step of the production process that is the layout. If the designs were otherwise, this composition would have been different: maybe Asuma wouldn’t be slightly hunched over to line up more naturally with Noa, and maybe the overall structure and sense of distance and height would be different. The comparison with a similar shot from Patlabor 1 clearly illustrates how all those parameters were considered in order to produce a far more detailed, convincing image. Moreover, as in cut 135, the way these designs are drawn by the animators or the animation director is essential. Here, there is no deformation, however small. In fact, if we follow Oshii’s line of thought, the consistency and “realism” of the designs and movement directly follows from his way to create compositions, that is from the layouts:
The deformation of the human body, which is an expression unique to animation, is fundamentally incompatible with compositions’ ability to produce an impact. For that reason, it should not be used too often or too carelessly.”
So far, I have reviewed the way Oshii himself understood the layouts and their purpose. But, as much control as he may have had over them, their creation was very much a collaborative process. Understanding how and by whom they were made will therefore further illuminate their importance in another way.
First, it is worth taking some time to discuss the differences between Oshii’s storyboards and the completed layouts. This was the result of multiple factors. The most general one is, as I’ve discussed above, the fact that Oshii started prioritizing layouts over storyboards. But, more precisely, this was also caused by Oshii’s methods and the production’s schedule: just like in Patlabor 1, the director insisted on thorough location scouting in Tokyo, which happened on multiple days between January and February 1993. Considering that the movie came out on August 7 of the same year, that left around 6 months to complete it – a timeframe similar to Patlabor 1’s. The design works and writing were completed at this point, and it’s possible that a rough version of the storyboards was was as well by January, photos were essential to the layout process, which had to proceed at breakneck pace, sometimes without definitive storyboards as reference.
Comparison between the storyboard, layout and final image for cut n°520. As we can see, the image on the billboard is different from storyboard to layout. It also appears that the model of Matsui’s car hadn’t been determined yet
This rush situation may explain why the number of layout artists is so high: whereas there were only two on the first movie, that number had risen to 6 on the second one – although those numbers remain quite low compared to Honneamise’s 20 and Rôjin Z’s 50-or-so layout artists. In any case, the profiles of Patlabor’s artists were quite diverse. The two veterans from the first movie, Kiyomi Tanaka and Takashi Watabe, were both mechanical designers, illustrators and concept artists. Masatsugu Arakawa, Atsushi Takeuchi and Yoshio Mizumura were IG animators whose work on the first movie had been appreciated by Watabe and who subsequently received a promotion. Finally, Satoshi Kon was a mangaka and shaping up to become a professional layout artist following his work on Rôjin Z, Run, Melos and Memories. To these 6, we should add animation director Kazuchika Kise, who corrected all layouts and often drew the characters in by himself when other artists didn’t have the time.
Watabe was the clear leader among the layout artists, handling the highest number of cuts: between 200 and 300, whereas other artists were responsible of around 100 each, on a movie which counted 872 cuts (an extremely low number – less than three times more than the average number of cuts for a TV episode, around 300). Watabe also had the privilege of making some of his layouts in 3DCG, something that Arakawa wanted to try out as well but couldn’t due to the budget being too low. While I will come back on the use of 3DCG in the film, using it to modelize layouts was certainly unique and prefigured the pipeline of Oshii’s future movies, notably Innocence. It seems that Watabe didn’t use CGI for Patlabor 1 (perhaps for lack of time and money), but he was probably familiar with the technique. Indeed, his close colleague Kiyomi Tanaka had been credited under both “Layout” and “3DCG Operator” on Honneamise in 1987. There, his work consisted of creating 3D models for planes and mechas that the animators would use as reference or directly trace. Perhaps even during Honneamise’s production, Tanaka must have become aware of the technique’s wider potential and, with Watabe, they probably started experimenting in that direction.
To conclude on the layouts, what all the above illustrates is that Patlabor 2 isn’t where a so-called “new” layout system was “invented”. Rather, it was the encounter between different visions of the layout system and what it should be used for.
The first one would be the “classical” view descending from the Takahata-Miyazaki layout system. There, the layout is understood as a blueprint for the animation – a preliminary form of rough animation. In that framework, the layout modelizes the movement of the characters, and it is this movement that creates a sense of space. Although Kise and the IG animators had no experience with such a layout system prior to Patlabor, their testimonies show that, like Miyazaki, they approached layouts like animators.
Then, we have the view defended by Watabe and possibly Tanaka. The complete opposite of the animators’ view, it considered the layout as a continuation of background setting and design. Therefore, space – the backgrounds – would exist prior to the characters, whose movement was entirely dictated by the predetermined space of the frame.
Finally, Oshii’s own perspective offered a synthesis to the two previous views. It conceived the layout as a simulator for the camera, and therefore introduced the fictional camera as a third, separate entity that would totally regulate the space of the screen from its ambiguous position as both inside and outside that space.
Patlabor 2, realism and objectivity
To sum up the previous section, Patlabor 2’s layout system was not just a specific production quirk, a way to uniformize animation or to make things more efficient. It was a given answer to given demands, partly historical and technical, and partly aesthetic. It is this third dimension that I would like to examine in the final part of this article: how the movie’s form was both the cause and product of certain technical decisions.
But then this raises the question of these decisions’ purpose: if we follow the general understanding of the layout system, the goal would have been to imitate live-action cinema. This dimension is often associated, if not outright confused, with the movie’s “realism”, but I hope to show that things are more complex. I believe that the one to have formulated this problem the most clearly was episode director Toshihiko Nishikubo:
I don’t think that either the staff or director Oshii are aiming for live-action, but rather that we are pursuing reality. Whether it ends up like a live-action movie or like something completely different is just one result. In other words, this production just happened to have a live-action style. […]
We make a film by asking ourselves how to make it look the most realistic. So if you take photographs and directly transfer them to animation, it will not look realistic. To make a film out of something, you need to eliminate elements and stress some others – doing this includes the animation and background art, and I think that the result is a movie that looks realistic. In fact, that’s what happens in live-action with the cutting and framing.”
Nishikubo’s point of view is worth taking into account because it makes a distinction between multiple concepts: “reality” and a “realist” presentation of it regardless of the medium; “film” as an overall medium, and “live-action” and, implicitly, animation, as two styles/techniques that are completely independent from aesthetic criteria. To this, we may even add Oshii’s own distinction in the definition of layouts between “movie” (映画) as a visual work of storytelling and entertainment, and “film” (フィルム) as a material object resulting from specific technical processes. In any case, Nishikubo’s argument is this: realism is the end, and imitating live-action only happens to be a means to reach it.
This distinction is very helpful, but I in fact disagree with Nishikubo’s interpretation in two ways. First, in my opinion, one would need to reverse the terms: live-action is the end, and realism is the means. But then this raises the question of why live-action should be imitated, which brings us to my second objection: that Patlabor 2 does not consider either reality or live-action to be an ultimate goal or standard, but rather aims to integrate as many kinds of images as it possibly can. This can be explained by an examination of the different production processes, from the layout to the final image.
First, we can discuss some of Oshii’s decisions as a director, and notably one which he describes as the creation of “visual noise”. By this, Oshii means that not all elements within a work absolutely need to be significant and convey some kind of information. As he says, “Some of the cuts that make up a film must not contribute to the development of the story at all. On the contrary, there must also be, so to speak, unintelligible cuts, sequences, and situations that disturb and confuse the story itself.”
Here, Oshii’s goal is clearly to dispute something one often hears about animation: that the artists control every little detail in the frame, and therefore that imagination is the only limit, that they can create meaning out of every little line if they choose. This is especially important when discussing a director like Oshii, often described as using a lot of visual symbols in his work. In fact, for Oshii himself, many of the elements that may appear as such are just “visual noise” meant to replicate what he perceives as an essential feature of live-action cinema and its own specific realism:
Whereas live-action productions easily include it in their own cuts (even if that is unintentional), the “background” and foundation of the worldview or, in other words, the depth of the world, is often missing from animated works. Just as it is the case for the real world, the reality of a film (its world) requires unnecessary elements and unclear information: noise. But it is clear that such noise is the hardest to depict in animation, in which “is only drawn what was planned to be drawn”. […]
How can animation, with its intrinsic limitations, give shape to the “depth of the world” and become cinema?
The “landscapes with birds” are an attempt to do so. The birds that appear throughout Patlabor 2 have no symbolic meaning whatsoever. To formulate it clearly, they are but noise that was inserted in the work’s “world” and “story”: they are nothing else than extraneous signs meant to disturb the system of symbols.”
In other words, Oshii seeked to go past the “closed” nature of animated works by inserting random elements that would mimic the unpredictable nature of daily reality, and how such unpredictable elements sometimes find their way into movies: shooting or lighting mistakes, a simple detail in the background, an actor starting to improvise… For Oshii, these are all positive elements, and, from that perspective, animation’s lack of boundaries becomes a limitation.
But this also means that adding “random” elements is bound to fail, because they will never be completely random. The most that this noise can do is to send the viewers on useless searches for meaning and cannot fundamentally trouble the movie’s unity from the outside. However, it can disturb the viewer’s attention: such is the case, for example, of the many birds that fill the screen during Shinobu and Tsuge’s final confrontation. If we agree that they have no specific “meaning”, their value is then purely aesthetic – they fill the frame and create particularly dense and beautiful compositions. The viewer’s attention will then be diverted towards those compositions, and eventually diverted from the drama that’s unfolding. In other words, the inherent beauty that can be found in a given shot is a double-edged sword: either to convey the specific mood of the story, or to create its own autonomous, closed world.
This approach to direction largely depends on Oshii’s conception of the image as “information”: depending on the purpose of a given shot, the layout can contain more or less information which itself transmits a certain amount of meaning. Such a view is also expressed in Patlabor 2’s designs and animation.
While it is often broken in TV anime because of time and resources constraints, an unspoken rule often invoked in animation is that characters should never stop moving for too long; were they to do so, the illusion of reality would be broken. This is something that realist animation in Japan had adopted as one of its principles, and theatrical works were the privileged site where such a philosophy of constant motion could bloom: both Akira and Gosenzosama Banbanzai stand as symbols of it.
Both Patlabor movies are in complete opposition to this: most of the time, the characters barely move and if they do, they do so extremely slowly and delicately. Paradoxically, this places an even heavier burden on the animators: what few movements there are are extremely important, and they are precisely the kind of thing that cannot be entirely planned at the layout stage. In other words, the layouts are extremely constraining for the animators because they do not allow for any idiosyncrasy, but instead direct the movement in a very specific, minute direction: one that aims to reproduce the slow and deliberate performance of real people – or perhaps of “real” actors.
But there, complete photorealism isn’t really the answer: as Nishikubo points out, something like rotoscope would have destroyed the illusion of reality. That slowing down the movement and reducing it to its barest elements feels more natural is an illusion of animated techniques. Another one of those “lies” can be found in Takada’s character designs: as I explained earlier, they are remarkably simple – much simpler than anime’s general designs and even than the character designs of other realist works of the time. As it becomes a lack of expression, this absence of detail makes us enter in the movie’s unique atmosphere, and therefore plays a key part in the spectator’s immersion.
To understand this better, we can turn once again to Oshii’s discussion of a specific moment: cut 494, described as “Gôto who has betrayed Shinobu’s trust”. Although very brief, this shot carries a lot of importance, and most of it is contained in the drawing of Gôto’s face. Oshii comments on it as follows:
A frontal shot of the face of a man with a guilty conscience is something that would have been unthinkable in animated works of the past. Unlike flesh-and-blood actors, animated characters are designed to express emotions with the movement of given facial expressions indicating joy, anger, sorrow, pleasure… The reason for that isn’t a question of drawing ability: rather, using such symbolical expressions is more convenient and rational when multiple people have to work on a single image. But, considering today’s general drawing ability and level of design, it is not impossible to create such a scene where the face does all the talking. Given the right situation and layouts, chances are that it might work.”
Then, we arrive at the elements pertaining more directly to the technical aspect of the layout system: the movie’s constant imitation of lenses and of a fictional camera being installed in a fictional movie set. This is where Nishikubo’s argument is the most fragile: if “reality” had been the ultimate goal, why go to such pains to reproduce artificial tools of perceptions such as lenses, sometimes very distorted ones like fisheye?
Besides the lens effects, Patlabor 2 frequently uses a specific technique to reproduce the impression of a real camera that exists in a real space: that is background animation. The ways in which it was used vary from scene-to-scene: sometimes, in a style very close to Akira, the background animation is supported by complex photography work and highlights the parallax effect created by the movement of all the different layers. Other times, it works entirely on its own as the entire screen is cel, which then puts the viewer directly in the character’s position.
While background animation is also prominently used in the movie’s action scenes, the overall goal isn’t to create a sense of excitement, as is generally the case with this technique. Or rather, if it is, it is an ambiguous kind of excitement: as we adopt the perspective of the weapons and their users, the feelings we feel may be the very same. In other words, background animation does not stand up for a movie camera, but rather for a weaponized perspective: that of people within a cockpit, or of the point of view of a missile. In that sense, it is true that Patlabor 2 does not just imitate live-action cinema: its ability to integrate different ways of seeing and representing the world is one of its core elements. But the question remains whether that is an element of “realism”.
To provide at least a partial answer, it is interesting to turn to another central element in the movie’s production: the use of 3DCG. It is largely in continuity with the way it was used in Patlabor 1 – that is, CGI is used to represent “real” CGI such as simulations, monitors, etc. But the scale was on another level. Between 1989 and 1993, the cost of CGI had decreased very fast, while the budget of Patlabor 2 was double that of the first movie’s. The 30-seconds CGI shot animated by one of Oshii’s friends on the first movie were now 66 cuts, amounting to around 7 minutes of footage, produced by post-production company Omnibus Japan and overseen by CG director Seiichi Tanaka.
Besides the higher amount of resources that were put into it, the CG animation process had become considerably more elaborate. In line with the close collaboration between divisions that the layout system established, the CGI sequences were the result of a dialogue between 3D animators and 2D background artists. Indeed, the former used texture mapping: they first did rough 3D models, to which they applied textures provided to them by the background artists. In some sequences, notably the missile speeding up to the Bay Bridge, 2D backgrounds were even used as a template. Here is how Tanaka explains the process:
For the CG of the bridge, we had the art director draw the background first. I asked him to create shadows and draw an outline. We then imported those, transferred them in 3D and moved them according to the camera work. […] We took the trouble of importing the drawings to make the texture closer to that of normal art. I wanted to reproduce the touch of a brush.”
Using CGI was a basic requirement for at least two reasons. One, the same as in Patlabor 1, was an issue of consistency: the technology of the fictional world and of the real world are the very same, and what the fiction has to say about technology therefore also applies to our real world. Two, more specific to Patlabor 2’s thematics, was to highlight the idea of “simulation” and of different perspectives – as I’ve said above, different ways of seeing and representing the world. But then, it is even more important that some of the CGI was somewhat hybrid and used 2D work as reference: that is because no system of representation (individual or collective, technical, ideological or aesthetic) is entirely pure: it always builds on multiple others by taking some of their elements and reframing them in another context.
Having said that, we can now turn to the photography work, whose results express the same idea. Patlabor 2 probably has one of the best photography in the history of cel animation. For this, we can of course credit photography director Akihiko Takahashi, who himself indicates one important factor for the movie’s excellence: the specific film it used. Patlabor 2 relied on a very sensitive kind of film that had just been issued by Eastman Kodak – according to Takahashi, it was the very first work in Japan to use it. Of course, this made things more complex as the photography staff wasn’t used to it and had to experiment – but it led to many new innovative expressions.
The film’s exceptional sensitivity notably meant one thing: that the photography staff could now stack more layers and filters in front of the camera without it feeling off or the lower layers being obscured. In Takahashi’s words, “this makes it possible to take pictures with almost no change in color even if the image is slightly underexposed, and it’s also easier to do special effects such as filter work and transmitted light.”
Filters played an essential role in the movie. Between cel layers, Takahashi would add panels of glass with certain chemicals in-between that would lower or even eliminate the impression that different objects are on different cels during the shooting. Such a technique would usually have obscured the overall image, but here, it only created a slight blur effect which actually served the movie’s atmosphere. Filters were also heavily used whenever reflections had to be represented – that is very often. In that regard, the most impressive moment is no doubt the shot below, when a helicopter flies by a building and is reflected on its windows. Layout artist Atsushi Takeuchi expressed how much constraints such shots represented:
Reflections are really hard. It’s simply twice as much work. You’re doubling the number of key frames, in-betweens and cels to draw and shoot. In the case of the shooting, you also have to consider all the different filters, so it’s probably more than double the work.”
Such shots are also, in their way, a specific kind of fiction. First, because they embody the idea of a war that is only fought and shown indirectly – through monitors or reflections on glass. But it is also because, in live-action, the shot above would be almost impossible: the weather conditions would need to be exceptional for a helicopter to reflect well enough to create a similar image, and you’d always face the problem of the camera being visible in the shot. In other words, here as well, animation-specific tools are used to go past the simple imitation of live-action cinema.
Oshii makes similar remarks about the use of the multiplane camera and its unique parallax effect, arguably one of the most “animation-esque” techniques that even Patlabor 2 couldn’t discard. Above, I quickly mentioned some Akira-like effects in the interaction between background animation and background art. However, the intentions are very different: Akira’s backgrounds and photography are used to create a sense of awe, of Neo-Tokyo as a towering, overpowering presence. The specific movement created by the parallax gave the city a sense of irreality and conveyed the impression of its almost transcendent existence. Patlabor 2 used the multiplane in a similar “un-cinematic” way, but to convey a different impression: that of a city whose existence is not monolithic, perhaps barely real. In other words, it strategically emphasized the contrast between cinematic and un-cinematic techniques in order to integrate as many modes of vision and existence as possible. Oshii explains this quite clearly in his commentary about cut 302:
This shot is composed of two things: in the foreground, an industrial complex which was drawn using reference photographs, and in the background, buildings that don’t exist in the real world. Since Patlabor 1, the fundamental idea for composition is, within each shot, to create contrasts between reality and fiction, between a past still full of presence and a future that glows like the reflection of the midday sun in the windows of the buildings. For such a scene, which forces the viewer to acknowledge the contrasting sense of presence between two contradictory worlds in the very same image, the most effective camerawork is a detailed use of the multiplane. But it must be remembered that setting the speed of the movement to be as slow as possible puts a heavy strain on the shooting.”
With this overview of Patlabor 2’s 2D and 3D animation and photography work complete, it is time to go back to Nishikubo’s argument and the relationship between animation, live-action and realism. For Nishikubo, realism was the end and imitating live-action was the means to that. His position is easier to defend now, since it is clear that live-action is not the only technique that Patlabor 2 imitates or integrates. The movie may appear to be dominantly “like live-action”, but it is much more.
However, this still doesn’t elucidate the nature of the “reality” that Nishikubo presents as the objective. As I’ve shown, when it needs to imitate live-action, Patlabor 2 lies, but it remains as true as possible when imitating simulations. Therefore, I would rather argue that Patlabor 2 focuses on the hybridity of all these images: none of them directly derive from some reality, but all lie at the encounter between other images and, more fundamentally, between reality and fiction. This is first and foremost the case of animation itself according to Oshii:
What does it mean to reproduce live-action images in animation? Asking this means considering the very foundation of animated film. Briefly put, “animated films” understood as dramatic movies that descend from “traditional animation” do not use painting as a standard for their creation, but rather originate in the memory left by live-action images. Moreover (and this is capital), live-action images themselves are fundamentally dependent on the physical (optical) properties of specific lenses, and absolutely do not derive from any kind of aesthetic concept or standard.
This is completely different from painting, where the human brain is involved and there is a direct connection between the image and canvas. At first glance, animators creating moving pictures may appear to do something similar, and the entire visual creation in animation to share a similar structure, but their essence is completely different. Animation drawing is first and foremost a step in a wider process, and for that reason does not exist entirely on its own. The entire course of animation production is planned backwards from the final image, and must be understood as the process of accumulating all the necessary parts. In such a proceeding, the “essence” that is felt (intuited) individually on each production site is just one of many phases.
Such a way of thinking may be difficult to accept for those who hold onto “traditional animation”, but is necessary in order to understand animation as cinema in the broadest sense of the word.”
To sum up this long and abstract quote, I would say that Oshii’s point is this: in animation, realism understood as a rendition of human perception and experience is fundamentally impossible. It is impossible because animation’s primary reference is live-action, which is itself not a realist technique. And it is impossible because animation is a collective endeavor that cannot, for that reason, reproduce any individual subjectivity.
But then, what happens when a movie takes advantage of these two factors? What happens when a movie intentionally imitates live-action, contrasts these images against others from various sources, and multiplies the production sites and vantage points inside and outside the fiction so that unity of technique and consciousness completely vanishes? Patlabor 2 was made as an answer to these questions. While the undeniable auteurist quirks and impressive animation consistency may make it seem like Patlabor 2 is an incredibly tight and cohesive movie, it is in fact the complete opposite: it consistently refuses to adopt just one point of view or mode of representation. But, precisely because it remains so consistent in its presentation, it avoids falling into pure subjectivism or audiovisual chaos.
This, in the end, is why the layout system exists and is so important to Oshii: not as a means of directorial control or as a magic recipe for realism, but as the thing that maintains the objectivity of the work. And this is the key to Patlabor 2’s so-called “realism”: not a rendition of individual experience or a representation of some reality, but rather a dominating, distant viewpoint that integrates all visions and images of reality without discriminating them based on any moral or aesthetic criteria. If we push this even further, we may argue that realism as such is an illusion, based on the illusion of a single, univocal reality. But paradoxically – and essentially – it is only through fiction that such a statement can be formulated and understood as true.
Books and articles
Bolton, Christopher. “The Mecha’s Blind Spot: Patlabor 2 and the Phenomenology of Anime”. Science Fiction Studies. Vol. 29, N°3, pp. 453-474. 2002.
Demnati, Alia. L’industrie de l’animation japonaise aux prises avec l’image numérique : la réponse d’Oshii Mamoru (1995-2004) (The Japanese Animation Industry and the Digital Image: Oshii Mamoru’s Answer (1995-2004)). Art et histoire de l’art. Université de Paris. 2020.
E_Hoba. “MAMORU OSHII book review [nonfiction] Part 07: METHODS FROM PATLABOR 2.” Manga/Anime Memorandum, 1597507381. https://ehoba.hatenablog.com/entry/2020/08/16/010301.
Fujitsu Ryôta. “『スカイ・クロラ』公開記念 押井マニア、知ったかぶり講座！第６回『機動警察パトレイバー２the Movie』というか、レイアウトの話” (Commemorating the Release of Sky Crawlers, A Complete Course on Mamoru Oshii! Episode 6: Patlabor 2, or Rather a Talk on Layouts). WEBアニメStyle, 08.08.2011. http://www.style.fm/as/13_special/oshii006.shtml.
Ruh, Brian. The Stray Dog of Anime : The Films of Mamoru Oshii. Palgrave Macmillan US. 2013.
Sampson, Thomas. “Watch the Most Realistic Movie Portrayal of Modern Aerial Combat Ever Made.” Task & Purpose, April 3, 2022. https://taskandpurpose.com/entertainment/realistic-aerial-combat-movie-patlabor-2/.
Kasai Osamu, ed. 機動警察パトレイバー 完全設定資料集 Vol.4 – 劇場編 2 (Patlabor: Perfect Establishment Data Vol.4 – The Movie 2). 一迅社. 2009.
Oshii Mamoru. Methods 押井守「パトレイバー2」演出ノート (Mamoru Oshii Methods: Director’s Notes from Patlabor 2). 角川書店. 1994.
Kise Kazuchika. “それ以前”の世界に『新しい攻殻』を作ること – 黄瀬和哉総監督の考える『攻殻機動隊ARISE』とは? (To create a “new Ghost in the Shell” in the world “before” – What does Kazuchika Kise, General Director, think of “Ghost in the Shell ARISE”?). マイナビニュース, June 21, 2013. https://news.mynavi.jp/article/20130621-arise/
———. 「攻殻機動隊」25周年リレーインタビュー 黄瀬和哉. (Relay interview with Kazuchika Kise on the 25th anniversary of “Ghost in the Shell”). アニメハック, October 30, 2015. https://anime.eiga.com/news/101482/
Ogura Hiromasa. 美術監督 小倉宏昌さん (Interview with Art Director Hiromasa Ogura). ちびねこトムの大冒険 公式webサイト. http://chibinekotom.com/tom_world/ogura/index_1.html
Okiura Hiroyuki. animator interview 沖浦啓之 (Animator Interview – Hiroyuki Okiura). WEBアニメStyle (WEB Animestyle) Production IG. May 26, 2004. http://www.style.fm/as/01_talk/okiura01.shtml
Oshii Mamoru. 押井守監督インタビュー 「『ミニパト』こそ究極の押井版『パトレイバー』」 (Mamoru Oshii Interview “Mini-Patlabor is the ultimate Oshii version of Patlabor”). 高畑勲宮崎駿作品研究所 Production IG. February 8, 2002. http://www.yk.rim.or.jp/~rst/rabo/news/oshii1.html
———. 押井守監督が自身の作品について語る！ 3週連続単独インタビュー 第一弾 ～パトレイバーシリーズ～—シネピック (Director Mamoru Oshii talks about his works! Three consecutive weeks of independent interviews Part 1 – Patlabor series). WOWOWOビジネス, March 14, 2018.