An artist develops a character and makes the first drawings. A director shapes the scene: where the character is, what happens to them, and how the audience will see it. An animator searches for movement—a turn of the head, a gesture, a pause—that gives the drawn character a personality. Work on the environment, lighting, and the assembly of the image follows. One specialist’s decisions become the starting point for another’s work; the scene is reviewed, revised, and gradually brought to completion.
Artificial intelligence is becoming more deeply embedded in everyday creative work, and its capabilities are changing before our eyes. Alongside these changes, we need to learn how to understand both the process and the result. Working with these technologies involves more than pressing a magic button or finding the right phrase. In professional production, it requires knowledge, repeated experimentation, and painstaking refinement. People learn the tools, test their capabilities, correct errors, and develop their skills every day. This work is the subject of our examination: how a finished scene is created, and what people working with new technologies actually do.
AI tools are now becoming part of this production process. We need to become familiar with them, understand their capabilities, and learn to describe their use accurately. Following a scene from the initial drawing to completion helps reveal where a system contributes, what it proposes, and how an artist continues working with its output.
The November announcements from Disney and its partners provide several concrete examples: concept art development, the preparation of 3D animation, and the creation of environments for a live actor.

What Disney and Its Partners Announced
On November 9, 2026, Adobe and The Walt Disney Company announced a multiyear strategic alliance. Adobe becomes Disney’s AI innovation partner, providing technologies including Firefly and Firefly Foundry. The announcement describes custom generative models tailored to Disney’s intellectual property. An example already in use transforms human-drawn sketches into fully rendered 2D concept art for theme park development.[1]
A separate event took place on November 3 at Disney Accelerator Demo Day. Disney presented its collaboration with Promise on hybrid filmmaking and Animaj’s work on FlipMove, an animation assistance workflow. FlipMove is already used on Ozzy Fox, a series Animaj co-developed with Disney Jr. The companies reported that a second season was in production.[2]
These announcements concern different divisions and partners. Walt Disney Imagineering develops experiences for parks, among other settings; Disney Jr. is involved in a particular children’s series; StudioLAB explores technologies for the studios. Applying information about one project to all Disney filmmaking would be inaccurate.
How a Sketch Becomes Concept Art
Disney discussed Imagineering’s collaboration with Adobe in June 2026. The company describes models customized using its own creative materials and an underlying technology built on proprietary and licensed assets. It refers to more than seventy years of Imagineering work, from concept art to technical designs. The stated aim is to develop ideas and explore alternatives more quickly.[3]
The word “concept” matters here. An image can help a team choose the appearance of a space, its mood, or a design direction. Its existence does not mean that a finished movie shot, construction plan, or final set has been produced. Adobe’s November release specifically confirms work on concept art and creative assets for park development, alongside broader research collaboration.
Imagine an artist drawing the silhouette of a fantastical pavilion and establishing its proportions. A model proposes options for lighting, materials, and surroundings, after which the artist selects and reworks the result. This is a hypothetical illustration of the described principle, rather than a reconstruction of a particular Disney project. Assessing the artist’s contribution would require seeing the original drawing, the system’s proposals, and the subsequent revisions. The finished file alone is insufficient.

What Happens to Artists’ Work on Ozzy Fox
Animaj’s published account of Ozzy Fox provides a useful animation example. The studio identifies Jennifer Oxley as the show’s creator and describes contributions from writers, storyboard artists, directors, and animators. Its production breakdown compares two episodes, Potty Party and The Floor is Lava. On the left is the approved animatic—a rough sequence of drawings with established timing. On the right is an animation playblast, a preliminary 3D preview before lighting, final rendering, and compositing.[4]
These comparisons are more informative than the label “an AI-made cartoon.” They show a transition between production stages. A preliminary 3D preview should not be mistaken for completed production: further stages remain. The studio’s breakdown, linked in the source note, contains the demonstrations and their explanations.
In November, Animaj presented FlipMove as a connected workflow: translating drawings into 3D poses, preparing scene layouts from an animatic, generating intermediate motion, and processing directors’ feedback. According to the company, cameras, character and prop placement, and poses remain editable. The system proposes changes that artists review and refine before animation approval.[5]
What the Model Receives and Returns to the Artist
Technical publications clarify what “generation” means in this setting. In its description of Sketch-to-Pose, Animaj explains that a model receives a drawing of a character and predicts values for the controls of its 3D model—for example, head rotation and hand positions. These controls form a character’s rig. The output is an editable pose, rather than simply a flat image.[6]
In an interview, Animaj researcher Anton Raël describes the next stage. An animator defines key poses, and a trained model proposes the motion between them. It uses learned patterns of movement for a particular character. The output returns to familiar tools such as Maya or Blender as rig controller values. This matters for practical control: the artist can change the proposed animation within the working scene.[7]
| Stage | Human contribution | Tool’s task |
|---|---|---|
| Story preparation | Script, storyboards, animatic approval | The described tools use prepared materials |
| Translating a drawing | An artist-defined character pose | Proposing an editable 3D pose |
| Scene layout | Creative intent and composition approval | Preparing cameras and the placement of objects and characters |
| Motion development | Key poses and performance decisions | Proposing intermediate poses and transitions |
| Revisions | Directors’ notes, manual refinement, and approval | Preparing editable revision proposals |
This table summarizes Animaj’s stated workflow. It does not establish the proportion of automation in each episode and is not an independent production audit.
In a hypothetical scene, a character notices danger and steps back. The artist can decide when the character freezes, where they look, and how they shift their weight. The system helps fill the motion between defined states. Yet a small change in a pause or body position can change the meaning: caution can become fear or a comic gesture. Editability therefore matters to a scene’s expressiveness.
Why Motion In-Betweening Is Creative Work Too
The authors of Adaptive Interpolation-Synthesis, a paper accepted at SIGGRAPH 2026, explicitly describe motion in-betweening as an artistically demanding stage in which expressiveness and rhythm are shaped. Their method combines learned interpolation with pose synthesis. The abstract reports a 3.5-fold speedup for in-betweening tasks when integrated into Maya.[8]
That is a result for particular tasks and research conditions. It does not mean that making any animated film automatically becomes 3.5 times faster. Nor does it establish an equivalent reduction in budget or team size.
In a separate publication, Animaj explains that intermediate poses shape acceleration and deceleration, anticipation, transitions, and the additional motion details that give a gesture life. The model attempts to learn these characteristics from existing animation data.[9]
I would therefore use the phrase “removing repetitive work” carefully. Repeated operations certainly exist. But some of the work being automated involves artistic decisions and professional skills. If a model has learned characteristic movement from earlier material, the people who created that material also belong in the discussion.
What Is Known About Backgrounds and Live Actors
In Promise’s demonstration at Disney Accelerator, a live actor performed in a rough “grey box” environment. The demonstration then showed how an artist could place that performance in different story environments. Disney describes this as an exploration of live-action production combined with generative workflows. The published account does not disclose the complete model stack, the amount of manual processing, or how tasks are divided between visual effects artists and generative systems.
There is a basis here for discussing environment creation. A technology demonstration, however, does not establish its use across all the studio’s films or the replacement of all background artists. The reviewed announcements contain no general decision that “from now on, AI creates every Disney background.” Distinguishing the established facts from an expanded retelling of the news matters.
How This Differs From Earlier Digital Tools
Automation at Disney predates today’s generative models. The studio’s account of Meander, for example, describes a digital drawing system used on Paperman and Feast. It combined vector and raster drawing features, could interpolate strokes between frames, and could carry drawn strokes along with a character’s underlying 3D animation.[10]
This does not make Meander a modern generative model. It demonstrates why saying “a computer helped create the image” reveals little about authorship by itself. The working method needs to be specified. Geometrically transforming an artist-defined line, predicting a pose from a drawing, and generating new visual details call for different explanations.
Human contributions are better assessed through observable actions: who created the source material, which expressive decisions remain in the result, what the system changed, and what the artist subsequently reworked. A software name does not answer those questions.
Creative Control and Working Conditions
A promise to preserve an artist’s control needs to be examined beyond the availability of editing controls. How much time does the team get to make revisions? Who can reject a model’s output? Do production targets, pay, and opportunities for junior artists to learn change? Public accounts of the reviewed projects do not provide complete answers.
The Animation Guild’s 2024–2027 collective bargaining agreement is instructive. It permits new technologies, including AI, and allows employers to require their use under certain conditions. It also preserves employees’ contractual rights, provides for advance written notice when generative AI use may be required, and allows consultation at an employee’s request when production schedules permit. These provisions apply to work covered by the agreement; they cannot automatically be extended to every Disney contractor or all Animaj employees.[11]
Such provisions do not guarantee that every job will be preserved. Assessing the consequences of adoption requires evidence about employment and actual production arrangements. Artistic control over a shot and an artist’s economic position are related, but they are different subjects of inquiry.
A Single Label Does Not Determine Authorship
In its report on AI and copyrightability, the U.S. Copyright Office distinguishes assistance to a person from a replacement for human creativity. It states that using an assistive tool does not, by itself, remove copyright protection from a human-authored work. Protection does not automatically extend to purely machine-generated material, however; the particular human contribution and control over expressive elements matter. According to the Office’s analysis, prompts alone to generally available generative systems are insufficient.[12]
This is the position of a U.S. agency, rather than a universal rule for every country. It also does not automatically resolve the separate question of whether particular training materials may lawfully be used.
For this discussion, the method of analysis matters. If an artist created a drawing, defined poses, changed the composition, and refined the animation, those actions need to be considered on their merits. The term “AI” does not describe them or settle rights in every element of the result in advance.
How AI Disclosure Actually Works
The proposition that “Disney will avoid a label while an independent creator receives one” cannot be treated as proven without comparable examples. We first need to establish which rules apply to each kind of image.
YouTube requires disclosure of realistic content substantially created or altered with generative AI. Its help page separately lists cases that generally do not require disclosure, including clearly unrealistic scenes, certain animation, minor edits, and some production assistance. This does not mean that all animation is exempt from labeling in every circumstance.[13]
In its May 27, 2026 announcement, YouTube also distinguishes where disclosure appears. Photorealistic, meaningfully altered content receives more prominent labeling near the player or over Shorts; information for unrealistic, animated, or slightly altered content may appear in the expanded description. The company also describes automatic detection of some generative content.[14]
The EU AI Act distinguishes obligations as well. Article 50 separately addresses providers’ machine-readable marking of generative system outputs and deployers’ disclosure of certain content. For deepfakes forming part of evidently artistic or fictional works, it provides for disclosure in an appropriate form that does not interfere with the experience of the work.[15]
European Commission guidance notes that generated backgrounds, special effects, and standard technical processing in filmmaking do not necessarily create a false impression of documentary authenticity. Assessment depends on the content and context. A demand that “every use of AI must produce the same visible notice over every work” does not accurately describe these rules.[16]
The absence of a prominent label on a particular cartoon may therefore have an explanation in the applicable rules. Establishing preferential treatment for a major studio would require demonstrating different treatment of genuinely comparable content. The reviewed material does not provide that comparison.
Disclosing Technology While Keeping People’s Names Visible
I believe mandatory labels should be supplemented with a clear account of production. Audiences need information about which elements a system proposed, which materials people created, and who was responsible for the final version. Ordinary professional credits should continue to identify specific roles.
For a hypothetical scene, a description might read: “Storyboards and key poses were created by artists. Intermediate motion was prepared with an AI tool, then edited and approved by the animator and director.” This is a proposed example, rather than a quotation from Disney credits. It supplements applicable platform and legal requirements; it does not replace them.
Technical provenance information can also help. The C2PA standard enables verifiable information about a digital asset’s history to be attached to it. Its own explanation emphasizes, however, that provenance alone does not establish whether the content is true. It should likewise not be treated as an automatic certificate of artistic quality or the full scope of copyright.[17]
Disney’s projects and those of its partners already provide material for a substantive discussion: we can identify a particular tool, its task, and the format of its output. For me, the next necessary step is to show human work with the same specificity. Keeping original drawings, scene versions, and revision histories does more to achieve that than arguing over a single word attached to a finished image.
People should remain visible wherever they have genuinely created a work. Information about AI use should help audiences understand how that work was made.

Sources and Notes
This article draws on public sources checked on November 10, 2026. Production descriptions come primarily from Disney, Adobe, and Animaj; their efficiency claims are distinguished from independent conclusions. Hypothetical examples are identified in the text. Platform rules and legal provisions are discussed in relation to the cited sources and the verification date.
Source notes and references
- [1]↩
Adobe. Adobe and The Walt Disney Company Announce Multi-Year Strategic Alliance to Support Disney Innovation with New AI-Powered Tools. November 9, 2026. https://news.adobe.com/news/2026/11/adobe-and-the-walt-disney-company-alliance
- [2]↩
The Walt Disney Company. Disney Accelerator’s 2026 Demo Day Explores New Frontiers in Storytelling, Robotics, and Guest Experiences. November 3, 2026. https://thewaltdisneycompany.com/news/accelerator-demo-day-2026/
- [3]↩
The Walt Disney Company. Why the World’s Leading Tech Companies Collaborate with Disney. June 18, 2026. https://thewaltdisneycompany.com/news/tech-companies-collaborate-storytelling-innovation/
- [4]↩
Animaj. Behind Ozzy Fox How Artists and Technology Came Together to Create a New Preschool Series. July 23, 2026. Includes comparisons of animatics and preliminary 3D animation for Potty Party and The Floor is Lava. https://www.animaj.com/post/behind-ozzy-fox-how-artists-technology-and-audience-insights-came-together-to-create-a-new-preschool-series
- [5]↩
Animaj. Animaj Unveils FlipMove An Artist-Led AI Animation Workflow. November 3, 2026. https://www.animaj.com/post/animaj-unveils-flipmove-artist-led-ai-animation-workflow
- [6]↩
Animaj. Sketch-to-Motion Workflow Part 1 of 2. November 12, 2024. Section on translating drawings into character controller values. https://www.animaj.com/post/animaj-sketch-to-motion-model
- [7]↩
Animaj. Inside Animaj’s SIGGRAPH 2026 Paper A Conversation with Anton Raël on AI Motion In-Betweening. May 5, 2026. https://www.animaj.com/post/inside-animajs-siggraph-2026-paper-anton-rael-on-ai-motion-in-betweening
- [8]↩
Anton Raël, Julien Boucher, Antoine Lhermitte. Adaptive Interpolation-Synthesis for Motion In-Betweening on Keyframe-Based Animation. 2026. arXiv preprint 2605.02742; accepted at SIGGRAPH 2026. The speedup concerns the investigated in-betweening tasks. https://arxiv.org/abs/2605.02742
- [9]↩
Animaj. Sketch-to-Motion Workflow Part 2 of 2. April 3, 2025. Section on the expressive qualities of intermediate motion. https://www.animaj.com/post/animaj-sketch-to-motion-workflow-part-2-2
- [10]↩
Walt Disney Animation Studios. Meander Animation Tool. July 2017; updated September 2019. https://disneyanimation.com/technology/meander-1/
- [11]↩
The Animation Guild. Master Collective Bargaining Agreement 2024–2027. Article 21 N, especially paragraph 3 b; printed pages 43–45. Contractual protections have a defined scope. https://animationguild.org/wp-content/uploads/2025/11/2024-2027-Master-CBA.pdf
- [12]↩
U.S. Copyright Office. Copyright and Artificial Intelligence Part 2 Copyrightability. January 2025. Conclusions on page iii; sections II C–F on assistive uses, prompts, human-authored inputs, and modifications. https://copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
- [13]↩
YouTube Help. Disclosing use of GenAI content. Current guidance; examples of required and non-required disclosure. https://support.google.com/youtube/answer/14328491?hl=en
- [14]↩
YouTube Official Blog. Improving AI labels for viewers and creators. May 27, 2026. https://blog.youtube/news-and-events/improving-ai-labels-viewers-creators/
- [15]↩
European Commission AI Act Service Desk. Article 50 Transparency obligations for providers and deployers of certain AI systems. Paragraphs 2, 4, and 5. https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50
- [16]↩
European Commission. Transparency obligations under Article 50 of the AI Act. Discussion of deepfakes, backgrounds, and visual effects in filmmaking. https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act
- [17]↩
C2PA. C2PA and Content Credentials Explainer. Version 2.4. Discussion of digital provenance and the limits of what it establishes. https://spec.c2pa.org/specifications/specifications/2.4/explainer/Explainer.html