Looking back, looking forward: Kanada’s late period

Cover image: a layout/concept art from Fullmetal Alchemist: Daughter of the Dusk by Yoshinori Kanada

This article is part of the History of the Kanada school series

It is generally held that Yoshinori Kanada went through two major shifts in the 1990’s, shifts that determined what the two last decades of his life and work would be like. In terms of style, there was the transition to an apparently radically limited kind of animation, with very irregular timings and a profusion of straight, geometrical shapes. In terms of career, he progressively retired from the anime industry proper to work in video games as an employee of Square (now Square Enix).

In my last article dedicated to Kanada, I began to explore the origins of his late style and tried to show that it was anything but a sudden event. I will continue this demonstration here by trying to show why Kanada adopted this style and what were his probable inspirations. I will also try to retrace in more detail the evolution of Kanada’s career. Although I very probably do not have access to all sources, the last ten years of Kanada’s life are the ones on which it is the most easy to find information, as the Internet was steadily growing and more testimonies exist from the time.

Let’s start where I left off last time, that is in 1992, after the production of Porco Rosso  and Download. On both those works, Kanada came in direct contact with the rising new generation of animators that emerged after Akira: on Porco Rosso, he must have met Mitsuo Iso and Shin’ya Ohira, while on Download, he worked alongside Tatsuyuki Tanaka and Kazuyoshi Yaginuma, two major animators from the realist circles. Besides them, he was probably aware of the way things were turning: he had worked on Akira, and probably seen Gosenzosama Banbanzai and The Hakkenden. Flow animation, which had developed on those two OVAs, was on the rise and even one of his closest students, Masahito Yamashita, was turning to realism. 

The industry’s working conditions were also changing. In a previous post, I hypothesized that Kanada stopped being a close Ghibli collaborator after Porco Rosso because of their new labour policy. The world of TV animation was also going through major evolutions: the number of animators per episode steadily rose, the second key animation system slowly made its way throughout more series, and the control of animation and episode directors was getting heavier. Basically, the TV anime world that Kanada had left in the early 80’s had almost completely disappeared, and there was no way for him to regain the kind of freedom he had enjoyed.

This context sheds light on the nature of Kanada’s work between 1992 and 1998: he worked almost exclusively on OVAs and openings of TV series. The little animation he did on actual TV anime episodes was often very short and not very strong – he most often made it to help out his students, but doesn’t seem to have been very interested in the material he was animating and only delivered subpar work.

The fundamentals of Kanada’s late style

It was therefore on movies and OVAs that Kanada mostly worked in the early 90’s. A good example of the kind of work he did would be the 1994 Yû Yû Hakushô movie, Poltergeist Report. I mentioned in a previous article that the TV series had been one of the most important steps in the transition between the Kanada style and flow animation. But it’s notable that, even though Atsushi Wakabayashi made a short contribution, the staff list of the movie had little in common with that of the TV series. Its director was Masakatsu Iijima, a Kanada student from the Z3 days, and it was most probably through him that the former could work on the movie.

Kanada was in charge of roughly a minute of animation, that is, relatively little (for comparison, it was a bit less than Shinsaku Kôzuma’s contribution to the movie), and not even the entire climax. It is, however, an important moment of the final fight, and Kanada seems to have given it his all. This sequence is quite representative of what he was going for at the time. The most remarkable aspects are in the layouts and choreography: they are very complex, especially during the camera movement/rotation from 0:37 to 0:40. Yusuke keeps jumping in all directions, and this accounts for the strong feeling of excitement that pervades throughout.

In terms of character motion, the return of straight lines is very obvious. Arms and legs are outstretched rather than bent, and the forms of the bodies are often very triangular. Combined with the comeback of speedlines and the simple, monochrome effects, you have something that’s both very dynamic and readable. The timing, as always heavily modulated, obviously plays a large part. In this sequence in particular, a strong feeling of density is added through the detailed debris animation. Kanada was no stranger to this, but here it reaches such a level that it’s hard to imagine that it wasn’t at least a bit influenced by Masami Obari, and maybe even by Takashi Nakamura’s work.

All the remarks I made here could also apply to Kanada’s other major work in 1994, his key animation on the last episode of the Final Fantasy: Legend of the Crystals OVA. Once again, he was called to work on it by old friends, its director Rintarô, and one of its animators, Takuo Noda. This is a decisive work, because it was Kanada’s first contact with the Final Fantasy franchise, and with game developer Square. It was most probably his interest in the universe he had to put into animation that contributed to push him, some years later, to join the company.

On the episode, Kanada animated part, if not most, of the final battle. His standout moment is this one-minute sequence from the climax. As a consequence of director Rintarô’s sensibility towards lightning and photography, the effects work is very close to the type Kanada had previously done on Hi no Tori: Hôô-hen eight years before, both in shapes and colors. But there are two very striking elements in the first half of this cut, striking because they were characteristic of Kanada’s early style from the 70’s, and largely (though not entirely) absent in his 80’s animation: cycles and multiplanarity. The lightning animation for the first fourteen seconds is just a series of cycles – it’s something very strong and obvious. As for multiplanarity, you can see it around 0:22: the beams and explosions are animated on a different cel to that used for the object they hit.

As for the character animation, the last twenty seconds of the sequence are very interesting. Generally, there’s a strong sense of rhythm based on an alternation of slow-ins and slow-outs with the pull-cels. What’s striking is how obvious they are: it’s not the kind of subtle modulation work Kanada or his students had done in the past, but something rather simple and in-your-face. This becomes even more striking when you go through it frame by frame: most of this part is on 1s or 2s. The timings were becoming less uneven, and it was in the very wide spacings that Kanada would create sudden shifts in motion – something called “snapping”, where the character’s position suddenly changes without any in-betweens even though the animation is on 1s. This technique had been initiated mostly by Masami Obari, and would be perfected by Hiroyuki Imaishi and become one of the staples of the neo-Kanada style in the late 90’s. Finally, the general approach to shapes and speedlines is exactly the same as in Poltergeist Report: triangular bodies that stand out against monochrome effects and straight lines.

To sum up, Kanada’s style in the early 90’s could be said to have been based on two relatively new principles: simplicity in the animation, and complexity in the layouts. He used the exact same tools and techniques he always had been using, but differently, with a sense for a more uneven rhythm, that stood out both against his previous animation and the fluidity of flow animation. In my last article about Kanada, I explained that by the early 90’s, his style could have gone in one of two directions: liquid, flowing animation, or rigid, crisp motion; it is clear that Kanada chose the second option, something that would become even more obvious in his following work.

Lucky Man and the return to A Pro

However, I believe that the most indicative change happened just a bit later, on the two openings Kanada did for Tottemo! Lucky Man. The show was directed by one of his earliest students, Osamu Nabeshima, and Kanada did, in addition to the openings, the character designs and some key animation on three episodes. The openings, especially the second one, are among Kanada’s most famous and iconic works, but they are important because they show that his search for simplicity and clarity only grew more pronounced as time went by. I’d go so far as to say that it drove Kanada to return to the fundamentals of his animation, and especially to one of his first inspirations, the A Pro style of animation.

One of the distinguishing aspects of the A Pro school is its minimalistic approach to layouts: its animators favored still shots and a frontal camera, which allowed them to display their talent for character acting without having to care for more complex kinds of motion. Even the layouts of Yoshiyuki Momose, who wasn’t from A Pro but worked alongside its animators, were simple and often featured a frontal camera. The second Lucky Man opening seems like a return to this philosophy in many aspects.

There remain standout moments of background animation, especially in the sequence from 0:30 to 0:38. But the motion into depth is less and less present; instead, Kanada favored simpler layouts and created a sense of space through multiplanarity rather than camera movement. For example, between 0:12 and 0:14, the different characters enter the frame one by one, probably on different cels, and most importantly laterally. The same applies to the slightly more complex fight and introduction sequence between 0:23 and 0:30. Compare this with the Galaxy Cyclone Braiger opening which features a similar kind of dramatic character introduction: here, the characters enter the frame frontally, and move into the depth of the image away from the camera.

What seems to have happened is that Kanada stopped focusing on the frame as a uniform, three-dimensional space. Just like the A Pro animators, he now considered the frame to be a sort of stage made up of different layers, each layer existing somewhat independently from the others and having its own individual motion to show off. The more radical approach to motion and timing in character animation is similarly inspired by 70’s animation, and especially Momose. In both cases, the timings are more uniform, mostly oscillating between 1s and 2s, with some 3s. It is the spacing and the poses that create the overwhelming sense of energy. Because of the intense use of snapping, it looks like every frame is a key frame, as a different pose is adopted each time.

Another striking 70’s inspired element is the approach to characters’ shapes. Despite their energy, Momose’s characters were often stiff, since there was little or no squash and stretch used and the body shapes were very simple and full of straight lines. Through the 70’s and 80’s, Kanada’s character animation had freed itself from that rigidity, something that would culminate in the liquid feeling of Birth, and in Download’s character animation. But already in his work from the late 80’s (such as in this scene from Laputa) Kanada had been trying to return to the energetic stiffness of the 70’s. The Luckyman OP represents the culmination of this search, but also demonstrates what still set Kanada apart: the idiosyncratic rhythm which avoids any feeling of repetition or any excessively mechanical motion.

It’s also on the opening  that Kanada’s renewed use of cycles was definitely cemented, as was his new, simpler approach to effects: in sharp contrast with his earlier liquid fire style, the lines would be sharp and straight, and the colors much simpler. It’s therefore very interesting to compare two apparently-similar scenes Kanada did thirteen years apart, both involving the famous “Kanada dragons”: the climaxes from Genma Taisen, in 1983, and from X/999, in 1996. As I mentioned in a previous article, the transition between the two is visible in Laputa: Castle of the Sky in 1986, but here I’m going to focus on the two extremes.

Genma Taisen’s dragons were a direct evolution of Kanada’s liquid fire; it is therefore no surprise that they are the visual representation of a volcano’s eruption. Their shapes are unstable and constantly evolving, coming close to morphing animation. This constantly-evolving quality rests on two major elements. First, there aren’t only dragons in this sequence: there are also birds that take shape and fly around the dragons, as if the lava didn’t adopt just one definite form. Second, the entire sequence relies on the use of color. There are three main colors: black, orange and yellow, and there is no clear demarcation between them. Some other animators (such as Shinsaku Kôzuma) would use thin black lines between each color, but this isn’t the case here. Each color therefore directly interacts with and contrasts against the others, creating an unpredictable, organic feeling. The number of curves is also notable: straight lines are reserved for the psychic attacks of the human characters, while the dragons themselves are mostly made of bent shapes.

Now let’s look at X. The most notable element is that now, the dragons are monochrome: they are only formed from different shades of red. The other important thing is that they aren’t born out of fire, but what looks more like beams or lightning. While their shapes are as unstable as the ones from Genma Taisen, this instability is expressed very differently: not through colors, but through lines. This sequence rests on the contrast between straight, lightning-like lines and sudden curves that seem to express the twists of the energy from which the dragons emerge. The other two big differences lie in the much more staccato rhythm of the X sequence, and the simplicity of its layouts: in Genma Taisen, the frame is crowded by the amount of color and morphing shapes, whereas in X, everything is easier to read and generally more flat. You can feel the multiplanarity and the heterogeneity between the dragons, the background, and the floating sphere, whereas Genma Taisen tends to make all these distinctions disappear in the free play of the colors.

In other words, Kanada’s animation had radically changed, and it had fewer things in common with his pre-Ghibli works than one might think. And this must have felt extremely fresh: after the extreme expressionism and sense of excess of Shin’ya Ohira and Masami Obari’s work in the late 80’s, one might have feared that Kanada would only double down on this tendency and fail to really innovate. But he went in the completely opposite direction, looking back and past his own previous work, thus renovating his animation from the ground up.

Kanada in Hawaii

X was, with Darkstalkers, Kanada’s last major contribution to the anime industry proper. It is somewhat ironic that his last great anime work, on Darkstalkers, was on a Capcom adaptation: in 1998, he officially joined the rival company Square and moved to Hawaii, thereby dissolving Studio Nonmaruto and severing many of his ties with Japanese animation. Not all his ties, however: during each of his trips to Japan he would make a small contribution to a movie, series or OVA one of his acquaintances was working on, and he also paid regular visits to the Comiket in Tôkyô. This departure from the industry was probably something of an event: for example, it was in 1998 that the book Yoshinori Kanada GREAT came out, as if to send him off, with illustrations and homages to Kanada from students, friends and colleagues ranging from Masami Obari to Leiji Matsumoto.

Kanada’s decision to leave Japan might seem sudden, but he was not the only one to turn his gaze towards the video game industry. In the late 80’s, the Kanada-style animator Hideki Tamura had left the anime industry to make his own games, and Shinsaku Kôzuma and Masahito Yamashita also joined Square at roughly the same time as Kanada, although it’s hard to tell if they joined on their own or to follow him. There were many reasons that would have attracted animators like them to Square: they were finding less and less freedom in anime proper, especially on TV series; they would be able to experiment with more resources and new technologies, especially digital animation and 3DCG; and, most importantly, the working conditions and pay promised to be better than in anime.

It seems like Kanada didn’t immediately go to Square. In 1997, his first video game work was the opening animation to the Sega Saturn version of the eroge Desire, developed by the company C’s Ware. His work there is anything but imaginative – it seems like he revisited his old techniques and iconic scenes in his new style. It’s also possible that what made him so conservative was caution: he was in a new environment, and didn’t want to experiment too much yet. His first credited work for Square was on the game Final Fantasy Tactics, in 1997. Kanada was credited as “monitor”, a term which seems to indicate Quality Assurance, that is testing the game to check for any glitches. This is a strange credit for him to have, and it’s most probable that he in fact did supervision work on the many animated sequences this version of the game included. His involvement in Square seems to have been relatively light at first: he is credited for three anime works in Japan in 1998 (Super Express Hikarian episode 52, Blue N°6 episode 1 and Alexander Senki episode 1) and was only given “special thanks” on Final Fantasy IX in 2000.

If Kanada’s workload on Square projects seems light between 1998 and 2001, it is because he was busy on the studio’s most important production at the time: the 2001 movie Final Fantasy: The Spirits Within. Taking four years to complete, this was a pioneering and monumental work, which made use of the most advanced processing capabilities for computer animation of the time, and aimed to revolutionize photorealistic animation through the use of motion-capture technology. All reports from the movie’s production indicate that Kanada seems to have been very respected by other Square employees. In their commentary track for the movie, the staff members always remark “this is Kanada’s animation” when one of his sequence comes up; more generally, many of the company’s members had grown on his work: in a talk about Kanada, Square executive Takashi Tomita mentioned Getter Robo and Nausicaä of the Valley of the Wind as the two works that left the biggest impact on him.

Kanada worked as layout artist on the film. He was accompanied by his students Shinsaku Kôzuma and Masahito Yamashita, who were both storyboarders. The layout work seems to have involved doing some sort of 2D animation which would then be rendered in 3DCG. Kanada was not yet involved in motion capture, as his layout work seems to have mostly been on action and effects scenes; this makes sense, as the realism the movie was aiming for in terms of character animation didn’t really suit his style.

According to the movie’s commentary track, some individual sequences took 3 months to complete – there was a lot of work put into it, but as impressive as it might have been back then, it clearly appears like a failure now. The character animation of the movie was stilted at times, but in my mind it is the effects work that suffered the most: all the explosions, beams and creatures have a strange texture which makes it impossible to really believe in them.

However, if Kanada felt any disappointment at the results, he wasn’t discouraged. On the contrary, he kept trying out new things: in 2002, on Final Fantasy XI, he would turn towards character animation as Motion Capture Animation Director on the game. In practical terms, this meant drawing original images in 2D, then directing the shooting of the motion capture sequence and supervising or modifying the final 3D version. From Tokita’s testimony, this was a long process, with a lot of trial and error, and Kanada’s last sustained attempt at directly intervening in 3DCG animation: after that, he would only work as storyboarder rather than animation director or layout artist. A notable exception to that is in the 2003 Hanjuku Hero 3D game, on which he did in-game animation of the titular character Eggman.

More significantly, however, Hanjuku Hero also represents Kanada’s return to 2D animation. He did the opening and ending animation for both the 2003 and 2005 games. The opening of the first one, Hanjuku 3D, is very interesting, especially because of all the 2D–3D interaction it features, and the stylistic dialogue it establishes with the earlier Luckyman opening. Indeed, the 2D-only sequences have a similar approach: very simple layouts, lateral entrances in the frame and a dynamic use of split-screens. However, in all the moments when 2D and 3D coexist, there’s a strong focus on the multiplanarity and the heterogeneity between the two techniques. Kanada uses it to its full potential, as the dynamic 3D backgrounds are the occasion for him to use complex layouts made of background animation and movement in depth. In terms of the motion proper, the strong poses and snapping are everywhere, and coupled with more modulation than what Kanada had been doing in his previous openings. With this, he showed that he had lost nothing of what had made his work so special in the first place: a tremendous sense of energy and fun.

Although the animation is very similar, the opening of the second game is very different, in that it involves no 3DCG but is full 2D animation. It was, in fact, an entirely solo work since Kanada did all the storyboards, backgrounds and animation by himself. In that sense, it is probably the most representative work of his late career, both in its animation and in the many parodies it makes: Kanada seems to have been appreciating more and more wacky references to anime and pop culture. He had always liked those, as his much earlier work on Daitarn 3 episode 22 demonstrates, but they had become a major feature of his work since his crazy, tokusatsu-inspired designs on Luckyman. The reference to Daitarn 3 is not a random one: in the openings of his late years, it seems like he had found the same kind of near-complete creative freedom he enjoyed back then.

Kanada’s last works

Things are more nuanced for his 3DCG work. Kanada’s last major contribution there was on Final Fantasy XIII, which was released in December 2009, just a few months after his death. He did storyboards on the game for events, cut scenes and summoning sequences. It is visible that, as years went by, he was gaining more and more mastery of his new tools. The Odin summon sequence, for example, features complex camerawork, gives off an unexpected super robot combination impression, and even has some form of Kanada poses. It’s nowhere near his 2D work, for the simple reason that the work on timing and modulation was impossible there – but it retains a characteristic sense of energy.

Besides this, it is symbolic that Kanada’s two last finished works in 2D were collaborations with some of his admirers. The first one was the opening for the Musashi Samurai Legend video game in 2005: Kanada did the storyboards, and the key animation was done by Hiroyuki Imaishi, whom Kanada explicitly asked to work with. One year later, Kanada would do, under a pseudonym, some key animation on the second opening of the TV show Gaiking: Legend of Daiku Maryû, on storyboards by Masami Obari.

The Imaishi/Kanada collaboration was a match made in heaven, as Imaishi had been following in the footsteps of all that Kanada had been doing since Luckyman. The storyboards were very stylized, with many simple pull-cels and highly kinetic scenes of action in-between. There, Imaishi showed that he had somewhat surpassed his master with even stronger key poses and snapping, but what I find to be the most interesting is the use of other techniques.

Kanada died 4 years after this opening came out, on July 21st, 2009, of a heart attack, as he was working on storyboards for the game Tactics Ogre: Wheel of Fortune. He was 57 and this was a sudden and premature death: as I mentioned, despite the relative setbacks of his 3DCG works, he was probably enjoying an unparalleled level of creative freedom, and it seems like he still had many projects in store. Among those, some are officially considered to be his “last drawings”, and warrant some analysis both because of their status, and because they offer a chance to talk a bit about Kanada as an illustrator: these are the artworks he made for the 2009 game Fullmetal Alchemist: Daughter of the Dusk.

This is just concept art, which explains how rough and off-model the drawings are. But precisely for this reason, they are very instructive evidence for Kanada’s style as an illustrator and his creative process. Here, everything is very indefinite and the image of the characters doesn’t seem to have been completely pinned down. The most interesting character here is Alphonse: in other drawings, you see Kanada has gone as far as possible on the steampunk aspects, adding pipes and chimneys everywhere. It’s still visible here as puffs of smoke go out of his helmet, but in this particular illustration, Kanada decided to go down on the samurai route: Alphonse has got the characteristic shoulder armor and the two swords of Japanese warriors. This shows the variety of his inspirations: among his unfinished projects, we find many characters inspired by Chinese and Japanese history and mythology.

This image is relatively dense, with the crowd in the foreground, the three main characters on what seems to be a stage, and various effects and objects flying around. However, it’s remarkably clear and readable, which can’t be said of all of Kanada’s sometimes very dense concept art. As I have many times already, it’s necessary to note the straight, almost rigid, aspect of every outline which makes the bodies adopt stark triangular positions. I mostly characterized Kanada’s stylistic evolution in terms of effects, but this is probably its greatest sign in terms of character animation and design. The characters from Birth and Download are all in curves, very malleable and open to deformation. His illustrations from the 70’s and 80’s are also full of little circular lines, as if the pen and the shapes themselves remained uncertain and not completely set down. Even in Kanada’s illustrations or rough key frames previous to the 90’s, there’s a lot of superfluous information added by a series of little circles or circular strokes.

Kanada’s drawing after Luckyman progressively departed from this. A remarkable step in this evolution is the ending of the 2003 Hanjuku Hero video game, mostly made up of stills by Kanada. The characters are simple, cartoony and childish, but in a very different vein from their soft, doll-like counterparts of the 80’s. Here, there’s definitely a picture book approach, with simpler, straighter and more rigid shapes. As in everything, he was taking the route of simplicity in design and techniques. This didn’t mean that his animation got more bare or less interesting: on the contrary, it enabled him to come back to his own fundamentals and refine them to the utmost.

Kanada’s death was a sudden and tragic one. There’s no telling where he might have gone or what he might have done had he stayed alive, and I’m not going to speculate on that. But if there’s one sure thing, it’s that he would probably never have stopped trying and exploring new things. His greatest strength as an artist was his endless creativity, and the amount of energy he put into discovering new ways to challenge his own art and the medium of animation at large. It is legitimate to mourn for all the works he could still have created and the paths he could have opened. But we can at least be happy about how much he had already given, and all the inspiration that had rippled out from his work to others.