The Aesthetics of Keyframe Animation: Labor,Early Development,and Peter Foldes

Celluloid animation and computer animation

In his textbook on computer animation programming, Rick Parent (2008: 44) explains the ways in which computer animation is generated:

In any case, animation is typically produced by one or more of the following: modifying the position and orientation of objects in world space over time; modifying the shape of objects over time; modifying display attributes of objects over time; transforming the observer position and orientation in world space over time; or some combination of these transformations.

According to this technical definition, computer animation can be understood as being generated through spatial modifications of objects or their viewing-displaying coordinates. With the definition in hand, Parent further categorizes computer animation into three types: artistic, data-driven, and procedural:

In considering computer animation techniques, there are basically three general approaches to motion control: the first is artistic animation in which the animator has the prime responsibility for crafting the motion. The foundation of artistic animation is interpolation …

The second is data-driven animation in which live motion is digitized and then mapped onto graphical objects. The primary technology for data-driven animation is referred to as motion capture …

The third is procedural animation, in which there is a computational model that is used to control the motion. Usually this is in the form of setting initial conditions for some type of physical or behavioral simulation (p. 2).

Parent’s categorization is rather representative of the current field of computer animation. While the application of keyframes is universal to all three types of computer animation, this article argues that the manipulation of keyframes derives from what Parent terms ‘artistic animation’. Parent’s category of ‘artistic animation’ is mostly adopted in the older medium of animated film, rather than special effects or the newer medium of game that rely more on motion capture and artificial intelligence (procedural animation).

Seen from the perspective of the computer, to animate means processing, via the algorithms, the value of spatial parameters set by the user. This is especially true with procedural animation, where sophisticated tweaking of configuration can be made with command lines. In this regard, keyframe animation appears to be merely another interface to modify the spatial coordinates of objects. Nonetheless, keyframe animation is not only the dominant interface for CGI programs today because of its intuitiveness, but was also the very core concept that informed the development of computer animation since its inception. In the late 1960s, a few computer scientists such as Takeo Miura and Nestor Burtnyk began to work on computer animation under the premise of alleviating the labor-intensive nature of celluloid animation. Picking up on the essential job differentiation between animator and inbetweener, computer scientists derived the concept for the computer to create the possibility of automatic inbetweening for animators. When transplanted to the computer, animators (keyframe setters) work on key drawings (keyframes) and leave the in-between drawings (intermediary frames) to the inbetweeners (to interpolate). This process was how inbetweening came to be replaced by interpolating. According to the Merriam-Webster Dictionary (2004: 654), interpolate means ‘to estimate values of (data or a function) between two known values’. Computer scientists thus devoted themselves to evolving better algorithms to ‘interpolate’ the keyframes set by the animators more efficiently and precisely.

What is a keyframe?

Lev Manovich, in his seminal essay ‘What is Digital Cinema’ (1995), rightly argues that animation as ‘a depository for nineteenth century moving image techniques left behind by cinema’ has now returned to the very heart of cinema production. If animation is indeed the new order of cinema, then keyframing as the foundation of moving images production clearly demonstrates just how the technique of animation has seeped into digital moving images. The practice of keyframing is not restricted to 3D animation systems such as Autodesk’s Maya and 3ds Max programs, but rather makes its way into the effects section in video editing programs, such as Sony Vegas or Apple’s Final Cut Pro, and even sound and music programs. Virtually everything from color and texture to cross-cutting effects, through camera works (crane, dolly, shutter speed, lens, etc.) and lighting and particles and dynamics (procedural animation of waterfall, smoke, fire, etc.) can be keyframed to define movement, properties and transformations over a specific span of time (frames). Also, in the post-production of motion capture, keyframing serves crucial vehicles, although complex keyframe manipulation can be automated, too. As such, every single aspect of the traditional cinematic apparatus and its effects can be configured in terms of keyframe animation. Therefore, it is not an overstatement to say that digital moving images are keyframe-based art.

As a basic example for keyframing, a car is keyed to be positioned at frame 1 near a tree, and again at frame 72 by a house in the distance. The computer will then automatically generate the in-between frames to produce this brief segment of animation, so the viewer is under the impression that the car moves from the tree to the house in three seconds. In addition to the setting of such positions, the animation program usually offers more options to fine-tune the keys in terms of velocity and momentum. In this way, the movement can be crafted to be convincing rather than robotic and dull. In the moving car example, the initial and last keys of the car can be configured with ease-in and ease-out settings. For more precise and nuanced control, effects and motion graphics programs like Adobe’s After Effects offer a field of ‘expressions’ for text-based programming expressions input, such as ‘wiggle’. Any sequence of motion graphics or computer animation is, therefore, made up of a chain of keyframes.

The effects of the keyframe system are most telling in the existence of the keys. Often, the over-emphasized role of the keys impresses motion or movement as being uncomfortably robotic, as if deriving from some lubricated mechanical joints, and it is annoyingly predictable. Certainly, as the computer, programming, and the art of computer animation evolve further, the phenomenon of glaringly cheap keyframing can be avoided. In this respect, Pixar has been exceptionally successful with its highly sophisticated computer animation work, such as Monsters, Inc. (Pete Doctor et al., 2001). Still, Pixar-style animation remains highly demanding in terms of both craft and time, regardless of computed automation. For the sake of efficiency, cheap keyframing in stock animation of video games or computer graphics for mass media broadcasting is today more common than not.

The division of labor

The first aspect about pre-computer keyframing is that the design was based on the division of labor between animator and assistant animator (inbetweener). In studio production, animators worked on the key drawings (extremes), determining timing and styles, while inbetweeners ‘filled in’ the in-between drawings. As we shall see, although the development of computer graphics was independent of cartoon studios, early developers were well aware of the division of labor practice. Specifically, the pioneering computer scientists emulated the animator/inbetweener split, and they began to devote themselves to algorithms that can automate inbetweening. Consequently, it is easy to see why keyframes always come in pairs, because the idea is for the computer to generate the intermediary frames on a frame-by-frame basis according to implemented algorithms.

The digital computer is indebted to its remote ancestor, the Analytic Machine, developed by Charles Babbage, who first dreamt of a machine to automate the toil of calculation. Babbage was also a keen industrialist. Deliberating on Adam Smith’s principle of the division of labor, Babbage suggested that productivity increased by repetition of limited movement or machine invention was ‘incomplete’(Hyman, 1982: 121). ² Instead, Babbage pointed out that the value of labor process fragmentation consisted of the discrimination into the cheaper or more expensive ones so that the capitalist can ‘purchase exactly that precise quantity’ of ‘different degrees of skill or of force’. Babbage’s idea of purchasing cost-effective labor inspired Frederick Winslow Taylor, who advocated the so-called ‘scientific management’ movement. In ‘Shop Management’, Taylor (1947[1903]: 94–100) advises that the work and duties of a ‘well rounded man’ should be divided into a number of positions because such a workman is difficult to train and expensive to hire. Also, in order for the system to work, Taylor insists the workman must never participate in planning and should be confined to limited routinized operations. While Taylorism increases productivity at minimum labor costs, Harry Braverman (1998[1974]: 78–81) criticizes its effects on labor alienation. Braverman extracts three principles from Taylorism: the ‘dissociation of the labor process from the skills of the workers’, the ‘divorce of conception from execution’, and the ‘task idea’. By depriving the laborers of any chance to determine or experiment in their working process, reducing them to mechanized repetition, and assigning trivialized tasks that prevent them from having any sense of fulfillment, scientific management cheapens and deskills labor.

Since computer graphics was founded on the idea of automatic inbetweening, I will now further expand on the early labor process of inbetweening. While inbetweening is a purely industrial concept, it is a fact that drawn animation evolves through the discerning process of differentiating certain drawings from others. As Winsor McCay realized in his vanguard experiment with animated cartoons, the work of animating can be simplified by what he called the ‘split system’. John Canemaker (1980: 20) writes:

Instead of animating an action ‘straight-ahead’ from, say, drawing 1 to 33, McCay would ‘split’ the action and draw first pose 1. He would then place another sheet of paper over the first drawing and make pose 33. On a third sheet of paper he would find the halfway pose, drawing 17 and continue to split the distance between drawings for the entire action.

Canemaker also noted that Winsor McCay diffused his method of the split system in his correspondence courses, but others who worked in the industry claimed to have already discovered this approach for timing. ³

It did not take long before early cartoon studios exploited the split system. Examining a photomontage from the Bray studio, Donald Crafton (1983[1982]: 165) notes there was Taylorist division of labor inside the Bray studio:

Instead of the old-fashioned, military-style shop organization, Taylor espoused ‘functional foremanship.’ This division of labor specified gang bosses, speed bosses, repair bosses, and ‘inspectors’ (teachers who demonstrated techniques to the workers). The individual department of the Bray studio corresponded to these functions … Management—Bray—is invisible (by himself on another page, actually). At the top are the artist-animators … Below them is a view of the camera room and a unique glimpse into the art department showing the ‘cheap men’ (and women) working on their daily tasks under the watchful eye of a female ‘inspector’ … The structure of the Bray studio (and most of the others) was pyramidal, with the founder at the top.

However, Crafton does not detail the job of these ‘cheap men’. Michael Barrier (1999: 20) describes the cheapening jobs revolving around celluloid processing. Some assistants traced pencil lines from paper to celluloid (the tracer), others blocked out the translucency with paint (the opaque). Therefore, the division of labor did not take place at the level of animating – until the birth of the inbetweener. Leslie Cabarga notes that the position of the inbetweener was created in the Fleischer studio in the 1920s. To enhance the productivity of Dick Huemer, the most talented animator at the studio at that time, the Fleischer Brothers proposed assigning him an assistant:

Max asked Dick Huemer if he would allow someone else to do his in-betweens so he could produce more work. At that time it was not only not done but quite unthinkable for an animator to let someone else touch his work. At first Huemer was opposed to the idea but in time his spontaneous revulsion gave way to his natural laziness and he agreed. (Cabarga, 1988: 36)

Ever since, animation drawings can be differentiated into degrees of value and importance, with inbetweening being an economically cheap part of animation production since an animator was more expensive to hire than an inbetweener.

The rather primitive labor rationalization in early animation studios eventually evolved to meet the demand of large-scale production at the Disney studio. By the mid-1930s, Walt Disney had developed an elaborate system beyond the simple division of labor between animator and inbetweener. According to Frank Thomas and Ollie Johnston (1984: 89):

An ideal group [for animation] would include an assistant animator who was experienced enough to make simple animation changes and corrections, a second assistant who drew well but was just learning his job, a reliable breakdown man, and an eager inbetweener who could double as bookkeeper and handyman.

Refining the earlier crude division between animator and inbetweener, the Disney studio further fragmented the full job of animator into animator, assistant animator and a second assistant, and divided the role of the inbetweener into breakdown man and inbetweener. While the breakdown man tackles ‘the main inbetweens that may contain special drawing problems’, the assistant animator, or more often referred to as the ‘clean-up man’, ‘changes and corrects’ (Barrier, 1999: 115).

Such an elaborate and intensive division of labor for the job of animation was essential to the expressive movement that Walt Disney sought to present in his studio’s work. In effect, this intensive labor rationalization mirrored the Disney aesthetics, which was only possible in full-fledged scale production. As Frank Thomas and Ollie Johnston (1984) explain, Disney animation encompassed such fundamental principles as Squash and Stretch, Moving Hold, Follow Through, Secondary Action, Overlapping Action, etc. The goal was to express the ‘thought process’ of the character through the intricate use of differential motion of various body parts and accessories, and thus the name of ‘Character Animation’. Among these components of complex motion, some parts will move at the same time, but at slower or faster speeds, while others will complete the movement much later. In The Mad Dog (Disney, 1932), the ears, flesh, and tail of Pluto are dangling, flapping, and whipping around while the torso moves through space and continues to show the aftershock of these movements. This exquisite calibration of motion will not be conspicuous to casual onlookers, but the resulting animation can definitely impress all viewers as being very ‘lively’ and sophisticated.

Seeing keyframe animation as inspired by the practice of celluloid animation begs the question of its dominance in the first place. In the animation industry, the medium of drawn animation allows the work of animation to be divided into style and timing design by the animator and execution by the inbetweener, so it is readily applicable to the division of labor. As Michael Frierson (1994: 64) rightly observes, for its resistance to labor rationalization, model animation was early on avoided by animation studios. Contrary to drawn animation, model animation would require a skillful animator to animate every pose before the camera. As each modification applied to the puppet is deemed equally valuable, so is the craft of the animator. Since the new automation was developed on the animator/inbetweener split, the inherent labor alienation was also carried over. The split in traditional animation may be hardly discernible, but once transplanted to computer animation, it becomes acutely noticeable: it speaks through the overly active keyframes and the tricky inbetweening.

Thus far, it is obvious that the development of keyframe animation consisted of the automation of the job of inbetweening. The underestimation was rooted in the conception of the division of labor that any labor process can be rationalized into the repetitive (the hand) and the more productive (the mind). Inbetweening was considered the less significant task that should be relegated to the lower echelons of animators while key poses were the precincts of the more valuable animators. Automatic inbetweening materializes the misrecognition of inbetweening in traditional animation. But to arrive at a thesis for the aesthetics of the keyframe animation, it requires an understanding of the mechanism for the automation of inbetweening. The following section will address the early development of deformation by linear interpolation, the primary method of automatic inbetweening.

Development of computer animation

Siggraph was the central organ that first promoted the dissemination of research in the computer graphics field, not the animation industry. Originally created as a Special Interest Committee On Graphics (Sicgraph) of ACM (Association for Computing Machinery) in 1963, Siggraph has served as a hub for computer graphics researchers from universities and the industry (Brown and Cunningham, 2007: 55). In the 1980s, as computer graphics leaped into three-dimensional imageries, Pixar Animation Studios premiered its early animated shorts at the Siggraph conferences.

The animation industry’s rather apathetic attitude toward computer graphics was partly due to the underdeveloped state of the early computer. Early computer animation experimentation was restricted to geometrical patterns and images, such as John Whitney’s Permutations (1966) or Poemfields (1966–1969), the series of mosaic-patterned computer animation by Stan VanDerBeek, who collaborated with the computer scientist, Kenneth C Knowlton. In the 1960s, the Canadian physician and mathematician, Leslie Mezei, attempted a more figurative kind of computer graphics. Although Frieder Nake (2005: 324) notes that Mezei used ‘geometric or stochastic transformations to line drawings’, his experiment only changed the dimensions of a single drawing or mathematically regrouped letters across the screen. Inspired by Mezei’s idea to transform mathematics into art, American artist/computer scientist, Charles Csuri, also designed some computer graphics pieces, notably Hummingbird (1967). This short piece won the prize for animation in the 4th International Experimental Film Festival in Brussels. Overall, these early works of figurative computer graphics were at best rotation, scaling, or translation (moving) of one line drawing picture. However, animation in the strictest sense rejects a fully automated process that only works to twist a drawing. As Gene Youngblood (1970: 203) rightly criticized of Hummingbird: ‘There’s no actual animation in the sense of separately-moving parts. Instead a static image of the bird is seen in various perspectives and at times is distorted by reversals of the polar coordinates.’

Figurative computer animation began with the keyframe system and the linear interpolation method in the late 1960s. One of the first interpolation algorithms was presented by Japanese researchers. Takeo Miura et al. (1967) proposed a new way of producing cartoons in a paper titled ‘An application of hybrid curve generation: Cartoon animation by electronic computers’. As they understood the process: ‘Motion picture cartoons and other types of animation require extraordinarily great expenditures of labor, since each individual cartoon frame must be drawn by hand’ (p. 141). To ease that workload, the researchers developed two ‘computing methods’. One method approximated modern vector animation in that curves could be mathematically produced with a series of parabolas, but it was limited to shapes made up of circles. The other method they proposed was ‘having two frames drawn by an animator. The curve indicating the movement between these two frames is then read into the computer’, so that ‘animations [sic] between the two frames are machine-drawn’. As Miura and his research fellows explained, since ‘interpolation is performed by the fundamental operations of movement, rotation, and expansion and contraction’, their method is called ‘linear interpolation’, upon which future development continued to build (p. 146).

Even though Miura et al. did not use the term ‘keyframe’ in the title of their paper, their idea of using a pair of frames as references for computation clearly originated in the cartoon studio practice of the split between the animator and inbetweener. Two Canadian researchers, Nestor Burtnyk and Marceli Wein, also picked up on the same idea, and they were among the first to use keyframes in paper titles. It was reported that Burtnyk’s idea of ‘keyframe’ animation came from a presentation by a Disney animator, which informed him of the division of labor between animator and inbetweener. ⁴ Burtnyk and other computer scientists easily grasped that a keyframe can be drawn by the animator, and intermediate frames may be interpolated by the computer. Burtnyk began to work on computer animation using the linear interpolation technique as early as 1969 in the National Research Council of Canada (NRC), which Wein joined soon after. In 1971, Burtnyk and Wein first materialized their research in ‘Computer generated key-frame animation’. At the same time, through the National Film Board of Canada (NFB), Burtnyk and Wein began to collaborate with the freelance animator, Peter Foldes. Their first film titled Metadata (1971) was minimalist in style, but consistent with Foldes’ apocalyptic themes about life, love, and death. The Canadian collaboration between computer scientists and animator continued and further crystallized in their next project, La Faim (Hunger, 1974). La Faim won several international film prizes, including the prestigious Prix du Jury at the Cannes Film Festival in 1974.

In a 1981 paper titled, ‘Inbetweening for computer animation utilizing moving point constraints’, William Reeves (1981: 263) assessed the existing development of automatic inbetweening: ‘The essence of the inbetweening problem’, Reeves noted, ‘is determining the correspondences between the keyframes.’ The algorithm pioneered by Miura, Burtnyk, Catmull and others in the 1970s, as Reeves pointed out, demanded specific instruction set by the animator for automatic inbetweening: ‘Each keyframe had to have the same number of curves and they had to be specially ordered by the animator.’ In his efforts to improve the existing algorithm, Reeves revealed the aesthetics created by the limitation of the linear interpolation of the seventies. Reeves argued that:

The classic keyframing interpolation example of automatically transforming a walking man into a speeding racing car actually illustrates what we feel has been one of the real bottlenecks of existing computer animation system—there is no control over the transformations, trajectories, and dynamics between the keyframes. (p. 269)

Reeves also noted that cubic interpolation has better control over the linear interpolation, but unfortunately it takes too much time for computation. Reeves’ solution was to provide ‘moving points’ and the surface patch to relieve the work of specifying correspondence and thus provide more control of the inbetweening process.

The technique of curve interpolation continues to evolve and eventually has been applied to shape interpolation in 3D computer animation. As Parent (2008: 148) points out: ‘Free-Form Deformation (FFD) is essentially a three-dimensional extension of Burtnyk’s technique that incorporates high-order interpolation.’ The 2D-to-3D extension is conceptually continuous because the space of computer graphics is natively three-dimensional. Early 2D animation inhabits a 2.5D space because the drawings are deformed locally and realigned with the world space axis. Parent credits the technique of FFD to Thomas W Sederberg’s 1986 Siggraph paper, in which Sederberg uses the bounding box, the lattice, around a solid geometric model to ‘sculpt’ objects or to distort them.(pp. 152–153).

By early 1990, the FFD system became the basis for ‘Animated Free-Form Deformation’ (AFFD), an important method used to produce 3D-computer animation. As Coquillart and Jancène (1991: 23) pointed out in their paper, it has been difficult to create a metamorphosis inbetweening that was not non-robotic (‘at the same time and speed’). Before their breakthrough, the keyframing for organic objects in 3D animation had to be almost manual, with lots of breakdowns and a large number of keyframes. With the AFFD method, an object can have animated deformation while traveling according to a set path and motion. In contemporary three-dimensional computer animation, deforming objects continues to be the staple for interpolation-based animation. ⁵

Timing and weight

In animation manuals, weight is a very useful concept to approach the rather obscure art of movement. John Lasseter (1987: 37) reminds his fellow computer animators: ‘No matter how well rendered a cannonball may be, it does not look like a cannonball if it does not behave like one when animated.’ The trick to animate a believable cannonball, Lasseter points out, lies in timing and weight: ‘The way an object behaves on the screen, the effect of weight that it gives, depends entirely on the spacing of the poses and not on the poses themselves.’ What Lasseter means by ‘the spacing of the poses’ is timing. As he tirelessly insists: ‘More than any other principle, timing defines the weight of an object. Two objects, identical in size and shape, can appear to be two vastly different weights by manipulating timing alone.’ In other words, timing can be understood as the artistic decisions on the spacing between poses, while the weight is the tangible effect of those decisions. For example, according to the principles of Disney animation, to move an object of great mass would require more effort in Anticipation, and its rigidity should show in the Squash and Stretch as it comes to a full stop and continues to show the reactions in the settling process. Therefore, all the Disney animation tenets such as Anticipation, Squash and Stretch, and Secondary Action, can be grasped as the tools to emphasize the expressivity of weight.

One of the advantages of Winsor McCay’s split system resides in its rather rational approach to timing. As opposed to the split system, the straight-ahead system offers a more intuitive way in which the animator improvises without much pre-planning. Obviously the split systems works better with cartoon studios. The advantage of the method of splitting, or what in Disney animation is referred to as the ‘pose-to-pose’ animation, consists of the fact that the design of timing can be doubled as a guideline for the division of labor. With celluloid animation, the studio system thus came to split the mind (design by animators) and the hand (execution by inbetweeners) of weight distribution and integrated the production into a collective effort: timing of the poses is constantly checked and reworked by a group of animators and their assistants on a drawing-by-drawing basis. In the case of computer animation, inbetweening is automated. The keyframes become the primary human labor anchors and thus their obtrusive existence.

Even though the concept of the weight remains central in Hollywood computer animation, the timing of movement would require the animator to grapple with the program. In computer animation, the design of the weight is primarily achieved by ‘keying’ the poses and by tweaking the interpolation. In other words, the gap between design and execution in traditional pre-digital animation studio seems to be finally bridged by the computer, which promotes each artist to the position of animator. However, computer animators now have a new struggle with interpolated frames. For instance, the Shrek case is symptomatic of the desperate attempt to mimic the weight in traditional animation with computer automation. However, Shrek also demonstrates how the tendency to deform adds to the difficulty for timing management, because animation by sculpting, distorting or deforming makes the expressivity of weight difficult. This is especially true of Metadata by Peter Foldes.

Peter Foldes

The art of the 20th Century is cinema. The language of the 20th Century is technology. In my films, I made metamorphosis. ( Peter Foldes, 1973, from personal communication with Bendazzi (Bendazzi, 1994: 433)

Peter Foldes was not a productive animator, but his works that include cut-out animation and computer animation received prestigious awards in his day. Foldes is now mostly remembered for La Faim in the contexts of the history of computer animation. The career of Peter Foldes as an independent animator is pertinent to the discussion of labor and the early development of computer animation. Like early computer scientists, Foldes was interested in the computer in order to save production labor. He never hesitated to point out how he felt plagued by ‘the inevitable slowness of repetitive work’ in animation filmmaking (Foldes, 1981[1972]: 204). The new technology of the computer for him was the perfect solution. His first computer animation piece Metadata was edited with the materials from two afternoon sessions when he first tested the computer system with the CNR scientists and technicians, while another animated short film Narcissus (1971) only took one day to complete. ⁶ More importantly, Foldes was acutely aware of the issue of economization of artistic labor because of his animation style, which extensively employed metamorphosis. In a personal communication with Giannalberto Bendazzi, Foldes expressed his appreciation of the computer when it came to metamorphosis that he was able to ‘work faster, because the machine frees the artist from the fatigue of labor’ (p. 433). Metamorphosis and labor were also catalytic to the birth of La Faim: according to Michael Century’s (2007: 86) documentation, Foldes submitted a drawn animation project to the NFB with a substantial amount of metamorphosis to transform one object into another. Century notes that because extensive metamorphosis would also require more assistants, Rêné Jodoin, the head of the French animation unit in NFB, considered the project ideally fit for computer experimentation.

Peter Foldes (a Hungarian) studied art at the Courtauld Institute of Art and the Slade Art School in London in 1946. As an art student, he was also interested in animated film. In London, Foldes learned the basics of animation from compatriot, John Halas, whose Halas & Batchelor Studio was considered the British Disney, producing the first British feature-length animated film, Animal Farm (1954). As he wrote in ‘The Computer and I’ (1972: 204), Foldes believed that motion pictures and paintings should come together, that ‘It seemed to me inevitable that in this age of motion, transformations, and continuous change, paintings have to move too. Motion pictures could be motion paintings.’ Soon, Foldes collaborated on his first animated short film Animated Genesis (1952) with his wife Joan, and won the Prix pour la couleur at the Cannes Film Festival and a Special Award at the British Film Academy awards. Originally coming from fine art painting, the Foldes’ animation was noted for its graphic expressiveness. Reviews in the newspapers reported that Animated Genesis ‘was a bold film composed entirely of mobile expressistic [sic] paintings’ (Russett and Starr, 1976: 203). His third film, A Short Vision (1956), also a collaboration, was screened on US television and due to its nuclear war topic caused quite a stir. ⁷

For more than a decade, Foldes went back to painting and did not work on animation. In 1956, he moved to Paris and began to work for ORTF (Office de Radiodiffusion-Télévision Française) in 1963. At the time, the ORTF spearheaded innovative computer art experimentation for both music and imagery; 1971 marked the first milestone for Foldes with two computer animation films: Metadata with the CNR, and the other, Narcissus, with the ORTF. Metadata received positive feedback from the Annecy Animation Film Festival, while Narcissus entered selection at the Cannes Film Festival. Metadata was an apocalyptic film of a man and wife from biblical Genesis to the end of civilization. Soon Foldes demonstrated to the world for the first time how computer animation could be achieved artistically with La Faim. This short film depicts how a slim and young businessman is driven by his insatiable appetite for food and is gradually transformed into a bold and fat old man. The obese businessman’s bulimic adventures end with his being devoured by famished children in the Third World.

However, the computer was not just a labor-saving tool for Foldes. Rather, he took advantage of linear interpolation in its primitive state and expanded his language of metamorphosis with technological gaucherie. As Parent (2008: 133) explains: ‘Because these [early keyframe] animation systems keep the hand-drawn strategy of interpolating two-dimensional shapes, the basic operation was that of interpolating one (possibly closed) curve into another (possibly closed) curve.’ More specifically, this is the ‘bottleneck’ that William Reeves (1981) discusses, namely, that early interpolation tended to fall into metamorphosis for its failure to fully control the interpolation process. By metamorphosis, Reeves meant the transformation from one object into another unexpected object. However, the most annoying defect was probably what Marc Levoy (1977: 65) described as the ‘collapsing phenomenon’, in which objects fell apart during the chaotic metamorphosis process. Surprisingly, for a great animator like Peter Foldes, this technical defect was a new device for creative exploitation. Foldes engaged with the tool at hand, transforming technical constraint into part of still larger artistic aspirations. La Faim is a case in point.

The aesthetics of early keyframe animation

Metadata was made with the concept of keyframe that Burtnyk and Wein implemented for Foldes. Foldes prepared the drawings with a light pen and set them as keyframes to be references for the computer work on metamorphosis. The effect of keyframing can be easily perceived here, as the movement always shows the gradual suspension of motion and changes of direction through its smooth transitions. As previously explained, pioneering computer scientists were adapting the computer to emulate studio inbetweener. However, just as inbetweening is now automated interpolation, the keys themselves become the highlight and where the aesthetics of automation becomes most ostentatious. In Metadata, the keys announce their own existence loudly. In the title sequence, the keys are again and again the end result of an unpredictable transformation, broadcasting the film title and credits. The film title ‘METADATA’ written in capital print is promptly metamorphosed into ‘Metadata’ as calligraphy then again into ‘un film de P. Foldes’. Metadata, as the title sequence visually suggests, is a film about metamorphosis accomplished by the automated computation machine. While the keys are intended to convey information, those in the male body sequence are employed to save artistic labor. The keys are set to make the body float, rotate, compress, and stretch in the abysmal dark universe. The presence of keys becomes very pronounced because they are points waiting for the body, as the body prepares to take on yet another corporeal transformation. Besides the presence of the keys, Metadata provides very little opportunity for weight expressivity, such as character walking or dancing. Perhaps the avid experimentation with metamorphoses was partly due to the technical potential offered and partly due to Foldes’ own artistic penchant. But obviously, the cosmic setting for Genesis, or the series of metamorphoses of the love story are ideal for weight-free animation.

One can imagine that the movement was new and intriguing to eyes untrained by computer graphics in the early 1970s. As the title sequence immediately demonstrates, computer animation is capable of an extremely smooth metamorphosis, almost like magic. The first scene of the film then begins with cell-like colorful circles floating in space, which then converge and metamorphose into a male body. Here, the rudimentary keyframe programming betrays its limitations. The body does not move, but metamorphoses from pose to pose. When the body metamorphoses from a huddling pose to a recumbent one, the in-between forms seem as unintelligible as barbed wire. When making a rotation in space, the simplistic sketch of the body only appears to be twisted and foreshortened or even simply abstruse. By comparison, the interpolation is more manageable in La Faim when the obese businessman falls into the Third World, showing a well-rounded mass of body fat tumbling into a gruesome abyss.

While these relatively unmanageable metamorphoses present themselves as typical of early algorithm development, the new computer medium provided an entirely new form of metamorphosis for Foldes. Like Metadata, La Faim avidly plays on the decomposing, shuffling and reconstructing of strokes to form a new picture. In fact, most of Metadata is dominated by such phenomena of disintegration and reconstruction. The assemblage of lines and strokes creates a woman from the male body, thus sketching the Biblical story of Adam and Eve. The first couple is then repetitively updated with a series of volatile exchanges of strokes, while switching to and back from a pair of hands. Eventually, the pattern of stroke recombination unfolds the daily life of the couple until the apocalyptical end of the city.

The reconstruction process during the metamorphosis differs from manual metamorphosis in that the transition presents an object as torn strokes, scattering and reconstructing in anarchy, which is utterly alien to the human. Interestingly, this ostensible novelty was actually inspired by a technical defect. Marc Levoy (1977) explained that such an effect belonged to one of the many drawbacks of early linear interpolation, and was only addressed by Burtnyk and Marceli (1976). As Levoy (1977: 65) points out: ‘linear interpolation between two substantially disparate drawings invariably results in the collapsing phenomenon, wherein the strokes of the drawing appear to disintegrate and mingle chaotically during the course of the transformation.’ By taking advantage of this collapsing phenomenon, Foldes creatively transformed a technical glitch into a new device for storytelling. Contrary to the emphasis on continuity in traditional metamorphosis, the collapsing phenomenon presents a rare glimpse into the progress of fragmentation and reconstruction. The artificiality of automated animation employed in La Faim, on the other hand, underlines the protagonist’s need to justify the reconstruction of environment according to his needs. The failed attempt to create a seamless metamorphosis betrays how automation naturally involves the destruction and reconstruction of our world. What happens in the reconstruction process is an impulse to modify the physical world according to the human interest, which is purely driven by greed for money, food and sex. In that sense, automation, if it feeds on greed, only facilitates the subjugation of Nature to merciless exploitation. Spaces are only precariously held together as the collapsing phenomenon warns of the destructive nature of over-consumption through the use of advanced technology.

Reflection on automation

Aided by computed inbetweening, Foldes also imbued metamorphosis with a reflection on human–technology relations, probably inspired by his own experiences with the computer. In La Faim, buttons served as the primary machine controls. Upon a button being pushed, the secretary emerges from a chair before his desk, carefully taking dictation. Also, the shop owner of the small grocery store that the businessman visits right after work has a cash register installed in the chest, which opens for change. What is relevant here is that buttons (along with a light pen for drawing) were the primary input devices for early computer graphics. While modern computer animation systems depend on the clicking and dragging of control points with a mouse, the early Canadian system assigned a group of points to corresponding buttons. In Foldes’ descriptions, the machinery he worked on was like a ‘cockpit’ for a pilot, installed with a control panel full of knobs and dials. Foldes explains the uses of buttons in the production process thus:

Set up at the command panel as if in the airplane cockpit (I must learn the role of each button by heart but it is like learning to drive a car), I can obtain all kinds of movements, colors, deformations, combinations of forms and background changes … I can determine the nature of movement, speed … and all the pre-established functions for different parts of a drawing. I engage the buttons to work on chosen combinations of points, a bit like musical synthesizer only it’s about movements. I can then accelerate or slow down, obtaining other deformations. (quoted in Roudevitch, 1973: 58, author’s translation) ⁸

As stated in Foldes’ account, the keyframe setting and manipulations were controlled through the buttons. Buttons in La Faim are, therefore, a self-reflexive reference to the machinery that facilitates animation making. But just as metamorphosis and rubber hose animation are automated, so was the society driven to obtain money and food. All this fluid transformation between human and tools suggests the human dependency upon daily machines is such that we grow into the technology we use. Human beings can thus be easily reduced to a specific function similar to a machine whenever necessary. After a brief session of dictation, the secretary has her head changed into a clock, switching back and forth to number 5 to inform the boss of the end of a day’s work, swiftly and automatically transmuting from a dictation machine to a clock. She can be summoned with the pressing of a button, very much similar to the ‘pre-established’ buttons on the control panel. The businessman himself is no stranger to this human–machine shapeshifting. In fact, the first thing the film reveals is that the businessman can split his head into two when he needs to speak simultaneously on two telephones. After work, the man is conveyed by the elevator downstairs and smoothly metamorphoses into a car on the street. Here, the interchangeability between tools and human makes a remote echo to the cheapened and deskilled ‘hands’ of an assembly line. How the interchangeable hands serve as machine parts of the conveyor belt predicts how humans will become a piece of machinery for the sake of efficiency and productivity one day.

Rubber hose animation

Computer keyframing greatly liberated Foldes from exorbitant labor, but it did not seem to impede him artistically. Crucial here was that Foldes did not attempt to suppress the inherent tendency of computer animation to deform, but rather incorporated it as part of his art. In fact, besides the extensive use of metamorphosis, La Faim also presents some free use of rubber hose animation – a variation of metamorphosis. A style popular in early film making, rubber hose animation allowed a free and arbitrary distortion of an object or a character’s body in their movements, and was generally employed to enhance hilarity in slapstick comedy acts.

In La Faim, rubber hose stands for an achievement as opposed to the collapsing phenomenon: the rubber hose sequence demonstrates that the animator can now successfully distort and transform curves into intended shapes. In effect, it was along the line of shape distortion that Burtnyk and Wein continued to develop their algorithm. They proposed some technical improvements after La Faim in ‘Computer animation of free form images’ at Siggraph (1975). Burtnyk and Marceli (1975: 78) believed that they now had better control over inbetweening with the ‘distortion or change of shape’. An illustration from Daphnis et Chloe by Foldes shows 10 frames of how an old man rejuvenates into a young man, in which the metamorphosis appears to improve significantly since La Faim. Unfortunately, their research did not gain further support from the NRC, and the project came to an end, but this paper became highly seminal in the field.

Sergei Eisenstein is probably the most renowned proponent of rubber hose animation, although he might not have been familiar with the term. He was exhilarated to find in Mickey Mouse something that shared his childhood fascination with amebic plasticity. The plasticity of the tubular limbs was in fact rubber hose animation, which Eisenstein (1988: 21) embraced as ‘the seemingly groundless scattering of extremities in Disney’s drawings’. ‘Plasmaticness’, as Eisenstein coined the term, celebrated the liberating euphoria rooted in prehistorical animism or the potentiality for an infinite number of forms.

Rubber hose animation originated with the American animated cartoon industry as early as the mid-1910s. According to Charles Solomon (1989: 29), Bill Nolan invented the rubber hose style, probably in the Bray studio. Originally, as Grim Natwick recalled, Nolan wanted to resolve the flickering phenomenon with ‘angular and straight’ characters in motion, so he tried ‘rounding the shapes and kind of swinging them into the motions’. The result, Soloman writes, turns ‘the characters’ limbs into lengths of flexible tubing: instead of bending stiffly at the elbow, an arm could move with enormous freedom’. The tubular character turned out to be capable of being more efficiently made while appearing to be funnier. Since then, cartoon characters have acted in their boneless existence, and the plasticity unique to the rubber hose style has proven to be an inexhaustible source for gags. Ub Iwerks was one of the most important animators to stylize the rubbery movement after Nolan. In Ub Iwerks’ ‘Oswald, The Lucky Rabbit’ series made before Mickey was ‘born’, trains, cars, and railroad tracks presented supple and elastic bends while Oswald nimbly distorted his body to absorb shocks. Later deemed crude and inappropriate, rubber hose declined in the Disney classical style by the early 1930s. Disney animators, under the aegis of more refined and uplifted taste, transformed the arbitrary corporeal distortion and anatomically incorrect tubular animation into the Squash and Stretch style we have become so familiar with.

Rubber hose animation continued to persist throughout the 1930s in the Fleischers’ cartoons, such as the ‘Popeye’ series. The rubbery style was indispensable to the eternal resilience of Popeye and Olive Oyl. However mercilessly Bluto launches his murderous attacks upon the couple, the plasticity of their limbs can always withstand inflicted forces. Rubber hose animation was also inspiring to young talent. Wilfred Jackson, a senior Disney director, once stated that his fascination with rubber hose was the reason for his lifelong devotion to the art of animation (Barrier, 1999: 73–74). ⁹ In an interview, Jackson gave an example of rubber hose that found an exact stylistic equivalent in the man’s lengthening arm in La Faim, namely, that the young man fetches his coat with his arm lengthening and expanding, not by walking toward it. Similarly, Foldes’ animated protagonist visits the grocery store by stretching and twisting his body; his upper body swings in while his feet remain on the pavement.

In her close reading on Pixar’s Wall-E, Vivian Sobchack (2009: 384) comments that automation is ‘informed’ by ‘an aesthetics of effortlessness, the lighter side of the alienation of labour’. Her remarks on early animation seem to suggest it had already anticipated the aesthetics of automation since the medium of animation is particularly prone to ‘sublimate’ its own laborious effort into a magic trick. Indeed, in the societies driven by the division of labor, the drudgery of routine works makes labor depressing. For Sobchack, the seeming lack of gravity is symptomatic of labor alienation. She comments that computer animation is ‘most often associating such “incredible lightness” not with (or as) ourselves but rather with (or as) moving image animation and its lack of gravity and with automation and its “little labour”’ (p. 391). Enlisting Eisenstein’s theory on the ‘plasmatic’, Sobchack argues that computer animation appeals to ‘the plasmatic’s seductive spontaneity, freedom and agency’ (p. 384). That is, the plasmatic provides an explanation for the ‘seductiveness’ of computer animation, which is deeply alienated because of automation.

Interestingly, it seems in the case of Disney animation, the more intense the division of labor on the job of animation, the more alienated the process, and the weight is more expressive. The chain of command from supervisor animator to animator and to the clean-up person makes sure the timing is perfect for the weight to shine through. In distinction to this, the computer animator is assisted by the automated inbetweening so that he or she can determine the timing of movement, and is thus more integral to the creative process. Meanwhile, the weight problem becomes the indelible mark left by automation, a carryover of the labor alienation of the inbetweener. I think the conundrum here lies in the fact that, as much as Disney animation has the most expressive weight, it does so with the aim to achieve the most delightful and comprehensible movement possible. Everything about character animation is for the audiences to effortlessly grasp the thought process and emotions of the characters through the art of animation. As such, Disney plasmaticness is prone to mask the intensity of its own labor alienation and appears so ‘effortless’ because the characters all seem so intimate and personal. In fact, the expressive style of Disney animation makes a perfect case for the critique of the culture industry. As Adorno and Horkheimer (1944: 137) suspect, amusement proffered by the culture industry has been inseparable from the effect of mechanization of work on the collective psyche:

Amusement under late capitalism is the prolongation of work. It is sought after as an escape from the mechanized work process, and to recruit strength in order to be able to cope with it again. But at the same time mechanization has such power over a man’s leisure and happiness, and so profoundly determines the manufacture of amusement goods, that his experiences are inevitably afterimages of the work process itself … No independent thinking must be expected from the audience: the product prescribes every reaction: not by its natural structure (which collapses under reflection), but by signals.

Therefore, per the critique by Adorno and Horkheimer, the popularity of Disney style animation may consist of the rather mechanic cues of reactions encoded in the movement that unconsciously comfort the audiences who share similar labor conditions. The effortless feel was also the implicit reason why Eisenstein (1988: 21) considered plasmatic animation an antidote to the ossification of Fordist society. Eisenstein contended that the popularity of Mickey Mouse’s ‘plasmation’ was due to its resistance against the ‘mercilessly standardized and mechanically measured existence’ of a Fordist society. The wavy movement of the boneless tubular characters was for Eisenstein a respite from the rationalization of labor process and its way of life.

Also, the seemingly effortless look as a representation of labor alienation should be considered in light of the absence of rubber hose animation and full metamorphosis in later Disney animation. As presented in Foldes’ La Faim, the fluidity in the complete changeover from one object into another or the extensive rubber hose distortion somehow underlines our confinement to detail work. Transgression into an inanimate state of being or to defy the pull of gravity is truly liberating, especially if measured against the rather momentary and transient Squash and Stretch deformation. More importantly, it was probably the dominance of Disney aesthetics that undermined the technical development of metamorphosis. No matter how mesmerizing metamorphosis or rubber hose animation in La Faim may seem, it is what William Reeves, one of the most talented computer programmers at Lucasfilm and later Pixar, saw as the ‘bottleneck’ of early interpolation due to its lack of control. Probably in Reeves’ view, the scene in La Faim when the businessman turns into a car would be a simple technical failure, if not claimed as being intended by the artist and scientists.

While overcoming technical difficulty is admirable, the furthering of the art of metamorphosis would be an aesthetic choice that would contradict Disney’s character animation. Changing a man’s head into a series of objects was never an ideal gag in Disney character animation. According to the principles of Disney animation, no matter how one character is distorted by Squash and Stretch exaggeration, he or she must be restored to the original mass and shape. This can be observed since early Disney to the more evolved character animation. Coming from the rubber hose tradition, Iwerks developed a fluid and plastic style informed by the silent animation tradition that has been unique to the medium of animation. Specifically, Iwerks transformed the old rubber style animation into a distortion-rebounding style. In Steamboat Willie (Disney, 1928), when Mickey’s belly is twisted like an extended rubber hose by the villain Peg Leg Pete, he restores his body shape by simply uncurling himself. In the 1930s, the exaggeration of Squash and Stretch further confirmed the importance of the maintaining of original form since it allowed partial transformation of the body that can be quickly returned to the original shape. Such restrictions on exploring form implicate a conformist priority in terms of social reality.

Metamorphosis continued to be developed, but not to be seen, in Hollywood computer animation. The lack of interest in the development of metamorphosis of 3D characters is said to be the taxing computation involved. However, the popular effects of fur and hair or shining metallic reflections (raytracing) are also extremely demanding effects for computation, and they are the subject of rigorous research. Still, metamorphoses are vital visual tropes. Notable examples are the haunting fluid metallic morphing of the terminator in Terminator II or the Michael Jackson’s ‘Black and White’ music video, in which racial boundaries are transcended by technological wonders. It is unfortunate that computer animation began with deformation and turned its back on it. As the Canadian government stopped supporting Burtnyk and Marceli’s pioneering development, and Foldes returned to oil painting, the fluid and arbitrary distortion and metamorphosis simply stops with La Faim.

Conclusion

The division of labor, as advocated by Adam Smith, Charles Babbage, and Frederick W Taylor, was invented to optimize the accumulation of surplus value. By fragmenting the labor process into individual and limited operations, jobs became easier to pick up so that deskilling could cheapen labor costs while attaining higher efficiency. If the division of labor was mechanized with Henry Ford’s conveyor belt and his assembly line developers, then automation would be the inevitable outcome since it is devised to further dispense with human labor. In this sense, the gliding motion of the conveyor belt already anticipated the interpolation between pairs of keys. The loss of authentic gravity by automated inbetweening therefore bespeaks the aesthetics of digital technologies, emblematic of a more rationalized capitalistic world that thrives upon higher efficiency and productivity. As this article aims to show, the weightless deformation that haunts the digital creatures and graphics discussed is the very manifestation of automation built upon the division of labor.

As automation replaces inbetweening, keyframe animation has been inscribed with the division of labor, and La Faim demonstrates just how far capitalist accumulation can go with automation as the most updated implement of labor rationalization. Nonetheless, instead of labor, La Faim addresses the context that gave rise to the division of labor, namely, the capitalist system. The film’s self-reflexivity of buttons and human–machine metamorphoses throws into relief how the interchangeability between human beings and new technologies is driven by faster money-making and consumption. Life is automated through buttons, much like the setting of keys to cue yet another change of action. This power of automation is most telling in the scene in which the businessman wraps up everything from the supermarket into his car trunk. Using a few keyframes, Foldes has the supermarket emptied and shrunk into a small box that automatically goes out of the frame as the businessman drives away smoothly. Taken this way, La Faim seems to portend automation with the automatic art of keyframing: how hassle-free technology only feeds the greed of the capitalist world.

While metamorphosis and free distortion can be highly liberating, Foldes was not as entirely optimistic as Eisenstein was about Disney animation. True, there is the utopian moment when the business man ‘swings by’ the gourmet store, but metamorphosis and free distortion can also be lent to the instrumentalization of mankind. It is probably the automated machinery that sets them apart. As the economization of artistic labor after the capitalist division of labor, keyframing is employed by Foldes to demonstrate how automation can be driven towards more relentless exploitation. Still, thanks to the embedded mechanism based on deformation, an unforeseen utopian dimension may survive in computer animation. As much as the contemporary deforming algorithm is highly amenable compared to that of the early 1970s, the unruly nature of rubbery distortion persists. Rubber hose animation was invented to work through the constraints of the cinematic apparatus put upon early animation and film, but it evolved into a popular style in American animation that fascinated not only the masses, but also critics. As computer animation comes full circle to rubber hose animation by way of deformation, it may be hard not to imagine automation being invested with similar utopian aspirations for the infinite possibility of form that animation as an art of movement so excels at producing.