Stop Motion Without Objects
Stop Motion Animation in the Age of Artificial Intelligence. Achieving Material in the Post-Material
From Generative Animation: Volume 4 of The New Machine Cinema
The puppet is gone. What comes after the puppet?
Stop-motion is the branch of animation most notably tied to its material constraints. Hand-drawn animation is tied to the craft of drawing. CG is tied to the render. Stop-motion’s identity revolves around its puppets: clay, foam, wire, wood, fabric; the entire tradition is a celebration of stuff. Both the puppet and the light upon it is real, while the film itself exists a live action condition. Yet no one says stop motion is not animation.
In fact, its live action condition is its trademark: the jitter between frames, the nostalgic trademark of stop motion, comes from the physical impossibility of placing the puppet in the same position after each adjustment.
Remove the object, and you remove the thing the entire tradition holds sacred. What I found, directing the stop-motion version of A Very Long Carriage Ride entirely made with AI, is that the removal of stuff produces its own kind of presence the traditional form could never generate on its own.
The Tradition of the Object
In 1912, Ladislas Starevich pioneered the medium by manipulating dead insects with wire armatures, producing miniature puppet dramas with real beetle corpses that moved with eerie precision. The medium of stop motion began as something like taxidermy in motion. This mythology is important to understand its origins: it began as a confrontation with the dead through reanimation. The point is, stop-motion is animation at its baseline, a literal artform meant for direct physical inhabitation.
Stop motion evolved from Starevich through Willis O’Brien’s King Kong (1933), and Ray Harryhausen’s dynamation creatures. It even achieved a formal life of its own as a kind of avant-garde, through Jan Švankmajer’s surrealist nightmares or the Brothers Quay’s fever-dreams. Mainstream comedies such as Aardman’s claymation, or Tim Burton and Henry Selick’s gothic spectacles, or Laika’s meticulous digital-hybrid features. The through-line across every category of stop motion cinema became the material through constraint.
The wobble of a puppet walking. The way clay moves across the shot. The way a miniature set carries actual depth of field because it is not being simulated from the flat plane of a drawing, instead, it exists in three dimensions being filmed by actual lenses no different than a live action film.
Even the way clothes move are subject to real gravity, just on a minute scale.
But material constraints come with material consequences.
This is why stop-motion inspires such loyalty among its practitioners and devotees. The form’s beauty is inseparable from its difficulty. The difficulty reads on screen.
Moving a puppet one increment at a time, photographing it then adjusting it, photographing it again, then again, across thousands of frames embeds a kind of earned fragility in every second of its running time. Therefore the difficulty of stop motion becomes part of the aesthetic contract between audience and creator.
The question with AI stop motion is whether the tradition of this contract is with the difficulty of stop motion, or with the result.
What the Latent Space Knows About Clay
The first thing I encountered on A Very Long Carriage Ride was an understanding of the exact same constraints traditional stop motion practitioners face. This on the surface is a contradiction. My stop motion came through images then animation, same as drawing, functioning as a two dimensional image on the screen. I could not go in and manipulate the puppets by hand. I had to do it through prompts.
Because neither the models nor the puppets existed. While the generative AI model knew exactly what stop-motion looked like.
In fact I found generative algorithms simulated stop motion animation almost trivially, capturing the same nostalgic motion of actual puppets in motion. This is important to note, in the entire animation continuum, 2D, anime, CGI; stop motion was mastered first by generative algorithms.
I recognized immediately between the two versions of A Very Long Carriage Ride, the stop motion was already operating against the highest aesthetic standard. The 2D animated version still had room to grow.
While directing, I estimated, if done traditionally with puppets and models, the stop motion version would have cost been $25-$50 million.
The long journey across Goosebury, the sophisticated carriage rides and locales, the cast of over 60 speaking parts; this was not a simple production regardless of medium.
The question was, how? While no company publishes what is actually going on underneath the hood, one can imagine the algorithm model was trained across the entire visual archive of stop-motion cinema: every Aardman film, every Laika production, every Švankmajer short, every student film and commercial and music video that ever used clay or puppets or miniature sets.
Because the compressed statistical residue stop motion on screen across every lighting condition, every surface treatment, every camera angle was simulated with such specificity, that I found no directorial limits. I could direct a $25-$50 million movie for $1,000 because my vision was in lockstep with the AI’s capability.
This is because the model gets these things right since they are visual patterns. They live in the image. And the latent space is an image-engine. Its competence is precisely the part of stop-motion that is visible.
Stop-motion is also physical. And here the encounter changes.
Weight, Memory, and Scale
The single largest gap between generated stop-motion and physical stop-motion is scale.
When a physical puppet stands on a miniature floor, it presses down. When it reaches for an object, its arm extends against the resistance of its own armature. When it walks, its feet make contact with the surface in a way governed by actual physics. The animator moves the puppet through space.
The micro-interactions between a character’s body and its environment, the exact way weight distributes through a puppet’s legs, the way a clay hand slightly deforms when it grips an object, the way fabric settles under its own weight between frames, these arrive with the same texture through AI simulation.
These are earned through painstaking physical actions.
What AI does is it gives the creator a direct relationship to product, instead of its process. Directing becomes abstract, almost second hand. It constantly feels as if you are one step removed from the process. The result is the same—that came first. But the feeling of creation becomes one of approximation.
The lessening of the relationship between creator and work is of note because the process of stop motion is the very artform of this relationship. Therefore the relationship itself must change. We must have a relationship to concept and scale instead of practice and effort.
Practice and effort remains, there is just an adjustment in how we conceived of the relationship between our actions and the result on screen.
Looking at the budget divide—$50 million to $1,000, one is also not limited by the physical restraints of the space within said $50 million production. Meaning you can have a scene with one hundred or one thousand puppet characters in motion, something not possible with traditional stop motion.
You can switch locations one hundred times, one thousand times. And I tried.
I felt it was important for the first AI stop motion feature film to achieve in a certain scale that would be difficult through traditional means, thereby achieving what a film such as Coraline could not, a film which operated as fundamentally small scale within a confined location.
There the benefits of machine maximalism emerged. The downside became the downside of AI itself. With infinite possibility available to you, have you done anything at all? This represents a kind of paradox with stop motion animation. Its entire purpose is to do; this becomes inverted, so the question becomes what is left? Can it even be called stop motion without painstaking labor?
The Nightmare and the Revelation
Can you be in love with the same artform if it operates through simulation and approximation? I found, yes. Because an interesting thing happened.
Alternating between the stop-motion and 2D animated versions of A Very Long Carriage Ride in ten-to-twenty-minute work sessions, I completed a sequence in 2D, then turned around and directed the same sequence in stop-motion.
Because I was producing the same film across two animation ontologies in parallel, I got to directly compare both processes in practice, scene by scene, cut by cut.
The first thing I found is, like stop motion itself, the AI stop motion was far more painstaking and difficult than its 2D animated counterpart. The effort when comparing the practices, traditional or not, led to the same result: painstaking difficulty.
In the instances of simulation and reality, you are attempting to inhabit inanimate puppets and bringing them to life. Through second hand approximation, the standard to do so is a lot higher because the bar of stop motion is already higher.
The 2D version was forgiving. It read as fantasy. When a drawing lives on a plane, and the audience meets it on that same flat plane, we use our imagination. We get swept away. And I had fun making it, finding it transformative and magical.
The stop-motion version was a nightmare, because stop-motion claims physical reality, and a claim of physical reality is far harder to sustain through generative synthesis than a claim of graphic abstraction. Because you are going through and adjusting the model through prompt, there are no do-overs for every take. A blown take requires another generation.
In its own way, the simulation becomes stop motion itself.
Every shot required a negotiation of labor. The performances required gestures, speech, expressions. Then I would study it against the adjacent frames and find that solving one moment of performance meant an entirely different moment was thrown off.
The director’s job remains the same: coherence and simplicity.
The sequence that taught me the most comes late in the film. The carriage carrying Autumn and her companions traverses a mountain road at night, hugging the edge of a vast cliff. The sequence cuts between the characters visible through the carriage window, the sheer drop beside them, and the waves crashing far below to establish the danger. It’s the film’s first Hitchcockian set piece.
After a jump scare, the passengers are forced out of the carriage, and suddenly they stand exposed on a windy cliff in the pitch black night. There is no carriage as shield, just figurines raw in the elements against an enormous landscape.
In the 2D version, this sequence worked almost immediately. The flat plane absorbs drama without resistance; graphic abstraction carries atmosphere through color, composition, and rhythm without needing to prove that the characters physically inhabit the space.
While the stop motion required real spatial depth to sell in the suspense. To achieve in compositional vertigo through the timeline and the scale. I found anchor frames where the contact between character and ground carried enough implied danger. When I built the sequence, cutting on the strongest frames, the stop motion feels genuinely immersive. It felt more real than its counterpart.
The stop-motion version was immensely more difficult and immensely more rewarding. It became my favorite version.
This was the revelation: directing generated stop-motion is a process of finding stop-motion inside the generation; the streamlining of the technology brings us full circle into requiring the same temporal coherence.
What Generated Stop-Motion Adds
The production of A Very Long Carriage Ride proved why stop motion is one of the most pleasurable forms of animation. The end result feels hand built in all cases. The pride of craft remains. The charm of viewing puppets and figurines carries over. One viewer remarked they wanted to take all the characters home with them.
Even the painstaking feeling of effort lives with you the same. The only thing that is missing is physical evidence.
Physical stop-motion is locked to its materials. A clay character is clay. A miniature set exists at one scale, under one lighting rig, in one configuration. Everything is committed at the moment of construction.
Generated stop-motion carries the same markers of materiality without inheriting the same constraints. Its limitlessness even applies to the camera, where camera motion can be simulated with great sophistication that would be restrictive or even impossible on a tiny stop motion set.
AI affords extra freedom in every direction. The only cost is the labor of thought.
These new confines in the artform become its own restraint: a lack of physicality. And its own aesthetic opportunity: the freedom of imagination. The end result is often beautiful. It is part of why stop-motion worlds feel so cozy, so self-contained, so much like dioramas you could reach into and touch. But stop-motion has always struggled with scale.
Open landscapes, vast architectures, magical forests; the sense of a world extending beyond the frame strain against the physical reality where the set is six feet across.
The environments in the stop-motion Carriage Ride have the texture of miniatures and the scope of landscapes. It is an adventure story after all. The world of Goosebury feels handmade and vast at the same time. That combination cannot exist in physical stop-motion because the handmade quality comes from its confines, its vastness requires a set that cannot be small.
The technology of the new paradigm begs for machine maximalism, otherwise, what’s the point of the new paradigm? It must be new, therefore it must justify its newness, therefore it must be pushed to its very limits.
The Drama of Materiality
Audiences respond to the drama of materiality. They have never responded to clay as innate to drama. But look the process of molding it, bringing ones hands in to the clay, squeezing out textures and form, decorating and dressing it with little clothes into fully form characters.
But the act of stop motion I posit is in the inhabitation, the same inhabitation of an actor meeting a great script on the stage. The spirit of performance must come through the reading.
The list of materials used across stop-motion’s history is enormous. Clay, foam, silicone. This rarely registers with a viewer. Same as generative algorithms, audiences have no idea what is going on underneath the hood, in neither generative algorithms, nor traditional stop motion animation.
So the drama of materiality itself becomes pronounced. The two versions of Carriage Ride, the 2D vs the stop motion, the simulated stop motion vs the physical stop motion, these versions all must compete on the very same level: the sense that something physical is being manipulated, that the world on screen has texture and weight, and that the motion carries the slight imperfection of real objects moving through real space. But most important, fundamentally, is that they perform.
That drama occurs first through a set of visual cues. The evidence lives in the image. But from there, AI can no longer help you. It has done its job to match aesthetics. What next? What becomes the purpose of the work?
What remains is the the same spirit of the animator regardless of form. The same considerations. The same constraints. Inhabitation. Performance.
The audience accepted CG when the physical drawing vanished. They accepted digital ink when the cel vanished. Each transition shed a material substrate that practitioners mourned, and each time the aesthetic encounter, the thing that actually moved the audience, survived the loss. Performance.
Stop-motion’s magic has always been that the puppet could be made to seem alive. It posited that ‘seeming’ is an image-event. And an image-event can originate from anywhere: a drawing, figurines on a tabletop, a CG render farm.
After the Object
Stop-motion without objects has crossed the same threshold every other branch of animation has crossed or will cross: the separation of the aesthetic from the method that originally produced it.
What survives the AI transition is the directorial encounter with the grammar. The drama of materiality. The tension between the handmade and the impossible. The warmth that comes from the imperfect. Rhythm becomes the quality of the image, belonging to the viewer’s final encounter with the film.
What is lost is also what is gained: constraint becomes opportunity instead.
The loss of constraint is so genuine that even this must be simulated and named. The True Line Cut. I imagine a digital simulacra with full 360 motion and interactivity. Touch will return. In this world is an endless study space with consequences; workers and a camera will exist in this digital space.
Even budgetary restraints return through algorithmic expenditures due to the new demands for scale; AI has lifted all standards for entertainment.
Because directors will wish to be directors. They will wish for hardship. They will wish for budgets, schedules, deadlines, rehearsals, and a final audience to witness the result.
When such opportunity are available to simulate with limitlessness, the simulation must encapsulate constraints as well. Constraints are what gives the endeavor meaning, ultimately. Stop motion proved that. AI disproved it, then will prove it again.
But for now, we are far from the point of meaning vs meaninglessness. The question of meaning in post-scarcity is a fallacy until abundance is truly actualized.
These questions accompany every technological transition in the medium’s history, in smaller and lighter forms. With AI, the question becomes total. Each time, the new form found its own resistances and produced things the old form could never have achieved.
The stop-motion version of A Very Long Carriage Ride differs from the stop-motion that Starevich or Harryhausen or Selick or Travis Knight would have made. It was made under the constraints of the latent space, constraints that produced a film carrying the warmth and tactile intimacy of stop-motion at a scale and scope no solo stop-motion production has ever achieved, at a cost of one thousand dollars, in a form that traditional stop-motion could never have afforded to exist in.
The object is gone. The drama remains. The drama was always what mattered.
