Short answer. Generative AI is already helping create storyboards, image variations, animation assets, and advertising materials. But the ability to generate a shot quickly does not, by itself, answer what story to tell, who makes the artistic decisions, or on what terms human labor may be used. For a studio, the sensible strategy is to test tools on limited tasks, keep the creative team accountable, and agree in advance on rights, consent, attribution, and compensation.
This piece is for producers, animation studios, marketing teams, and artists who are setting rules for working with generative AI. It analyzes the main points of Jeffrey Katzenberg’s essay “The World is Changing: AI For Creativity,” provided to the editors by a user, and compares them with historical and official sources. This is an analytical paraphrase with original conclusions, not a translation or publication of Katzenberg’s text. The legal treatment of any specific project should be assessed separately under the applicable law and contracts.
Prepared for AgentSunrise on September 24, 2026. Facts have been verified against the listed primary sources as of that date; industry forecasts and recommendations are separated from confirmed events.
Contents
- What Hollywood and tech companies are arguing about
- The history of technology in animation
- Why the historical analogy does not give a simple answer
- What AI can do in the creative process
- What conditions creators and studios need
- How to run a pilot without losing control
- Frequently asked questions
- How AgentSunrise can help with a pilot
- Conclusion
What Hollywood and tech companies are arguing about
In Katzenberg’s essay, he describes two reactions to modern video made with AI. He recalls his own impression of Pixar’s short film “Luxo Jr.”; a friend who is an artist asks whether what she saw means the end of her profession. These reactions are compatible: technology can expand artistic possibilities and at the same time change demand for specific kinds of labor.
Katzenberg argues that we should look first at the terms of use of the technology. Model makers get a powerful tool, but they need the stories, visual language, and experience of the people on whose work the industry grew. For artists and studios, a vague promise of “new opportunities” is not enough: they need clear decisions about source materials, consent, attribution, and payment.
In 2023, Katzenberg predicted that new tools could reduce the time and cost of producing top-tier animation “by up to 90%” within three years. That is the author’s own forecast, not measured savings from a typical project. It cannot be transferred to a studio budget without comparing the full cycle: story development, revisions, quality control, compute, legal review, and delivery. Source of the claim: the essay provided by the user; the 2023 speaking appearance is reported by Cartoon Brew.
The history of technology in animation
Katzenberg’s historical argument is strongest when it shows two things at once: a new tool changes the visual language, and the transition can take away people’s familiar work. If you leave out the first part, the conversation with artists becomes unfair.
From the phonograph to sound in film
Composer John Philip Sousa criticized mechanical music in the early 20th century. His concern was not only aesthetic: works were being reproduced by technical means, while the rules for compensation for such use were still taking shape. The Library of Congress describes his role in the debate, and the U.S. Copyright Office confirms that the 1909 law brought mechanical reproduction of musical works into the copyright system. This is an example of negotiations over the value of using work, not proof that every new technology is automatically good for creators. Library of Congress, U.S. Copyright Office.
The arrival of recorded sound in film created new creative professions and changed the audience experience. But live musical accompaniment in movie theaters stopped being the same source of employment. So the claim that “the industry grew” does not answer the question for an individual professional: can they move into a new role, and on what terms? The historical analogy helps frame the question, but it does not guarantee a new job.
CAPS, Pixar, and a new visual language
Disney and Pixar’s collaboration on Computer Animation Production System (CAPS) began in 1986, according to Pixar’s history. The system helped move a number of traditional animation steps into a digital workflow. Disney separately notes its role in creating the computer environment for the ballroom sequence in “Beauty and the Beast.” The key point here is not the slogan “the computer beat the paintbrush,” but the link between the tool and the artistic goal: the camera could move through space in ways the previous production process could not support with the same freedom. Pixar history, Disney material on the ballroom sequence.
In 1995, “Toy Story” was released, the first feature-length computer-animated film. The technical milestone mattered because the film worked as a story with characters, conflict, and emotion. Pixels alone did not create audience attachment. Pixar timeline.
| Transition | What expanded | What needed protection |
|---|---|---|
| Mechanical reproduction of music | Distribution of recordings | Compensation for use of compositions |
| Recorded sound in film | Musical films, sound editing, sound design | Work for musicians employed in silent-film exhibition |
| CAPS and computer animation | Shot space, compositing, new production techniques | Retraining and employment of workers from the previous process |
| Generative AI | Fast exploration of visual and story options | Rights to source materials, consent, compensation, creative control |
Why the historical analogy does not give a simple answer
Generative AI differs from CAPS in an important way. CAPS was an image production system; a generative model creates new material based on patterns extracted from training data. That raises separate questions about where the data came from, the use of a person’s image and voice, how closely the output resembles existing works, and who has the right to control the final work.
In the United States, the Copyright Office looks separately at digital replicas of a person, the copyrightability of outputs, and model training. In its report on copyrightability, it explains that AI assistance does not by itself eliminate protection for a human work, and that a prompt alone may not be enough to establish authorship of the result. This is the position of a U.S. agency, not a universal rule for every country. U.S. Copyright Office: research overview, report Part 2 announcement.
The World Intellectual Property Organization also warns independent filmmakers that using generative content may affect third-party rights, including copyright and rights of publicity. For a studio, that means it is necessary to review every source asset and agreement, rather than relying on the general label “made by AI.” WIPO guide to rights clearance for filmmakers.
There is also an economic difference. AI can reduce the cost of creating a frame variation, but project cost is not the same as generation cost. Developing the story, directing, fixing mismatches between scenes, coordinating with rights holders, and checking final quality remain separate work. If a test measures only generation time, the productivity conclusion will be overstated.
What AI can do in the creative process
Generative AI is useful as a tool for exploring options. It can quickly show several color directions, prepare draft references, help with storyboards, and suggest editing or ad versions. These uses differ in risk: an internal concept sketch for discussion and a finished frame in a commercial film require different levels of review.
It is important to distinguish three levels of work:
- Research. The team tests shapes, angles, palettes, and mood. The outputs do not go into release without separate selection.
- Production draft. AI helps prepare a foundation that the artist or director substantially revises and approves.
- Published asset. The result goes through artistic, technical, and legal review, and the terms of use for source materials and performers are documented.
This is exactly the transition between these levels that is often lost in the discussion. A convincing short demo shows what the tool can do; it does not show the cost of a full film production workflow or confirm that the rights to the materials are clear.
Katzenberg contrasts the model's analytical ability with human taste, intent, and empathy. That is a useful description of his position, but not a scientifically proven boundary that cannot be crossed. In practice, a more important question is who, in a specific project, defined the vision, chose among the options, fixed the errors, and is accountable to the audience and production participants. The model may provide an option; responsibility for the decision stays with people and organizations.
Which tasks are best to trust the tool with first
For a first experiment, tasks work best when the result is easy to evaluate and the harm from a bad option is limited: searching for composition for internal discussion, palette testing, a rough edit of a commercial, or adapting already approved materials for different formats. The closer the result is to a recognizable person, someone else's style, a licensed character, or the final shot in a film, the higher the requirements for approval and review.
| Task | Benefit of the pilot | Main check |
|---|---|---|
| Internal visual references | Quickly compare directions | Do not use the result as a finished third-party asset |
| Rough storyboard | Discuss rhythm and composition | Preserve the director's and artist's role in the selection |
| Ad creative variations | Test different approaches | Verify rights to the brand, source materials, and images |
| Final shot or character voice | Potentially reduce separate steps | Confirm consent, licenses, quality, and contract terms |
What terms do creators and studios need
Artists' demands cannot be reduced to fear of a new tool. Four separate questions matter: which works entered the process, who authorized their use, how the human contribution is recognized and who gets paid. Technological progress does not remove any of them.
The Writers Guild of America (WGA) says that the 2023 agreement does not recognize AI as the author of script material under the contract, allows a writer to use the tool with company approval, and requires that writers be informed about materials provided to them that were created by AI. The guild separately maintains its position on the use of writers' work to train models. This is an example of contractual terms for a specific market and worker category, not a ready-made template for any Russian project. WGA clarifications.
The actors' union SAG-AFTRA says that the 2023 agreement included provisions on digital replicas and AI; in 2026, the union reported additional terms for synthetic performers. The scope of specific rights depends on the text of the applicable agreement and the situation. The main management takeaway: a person's voice, face, and distinctive performance cannot be treated as a routine anonymous input file. SAG-AFTRA materials on AI.
For the team, it is useful to document rules before launching the pilot:
- a list of permitted source materials and the people responsible for confirming rights;
- whether consent is required to use voice, likeness, and performance;
- the storage rules for uploaded scripts, frames, and unpublished materials;
- the role of each creator and the payment terms for additional uses;
- the process for checking similarity to existing characters and works;
- the artist's or director's right to reject the result and return the task to the usual workflow.
This is a working checklist for discussion with the team and legal counsel, not a statement that one contract clause works in every jurisdiction.
How to run a pilot without losing control
An editorial model for a first experiment is "task → rights → creative choice → review". It connects the technical test with how people work. The model is proposed in this article as a practical way to organize a project; its effectiveness has not been confirmed by a separate study.
1. Choose one limited task
For example, creating several draft background options for an internal storyboard. Define right away what counts as success: the number of usable options after artist review, the time to an approved decision, and the number of revision rounds. Do not compare generation speed with the full timeline of the former team.
2. Check the input materials
Make a list of uploaded frames, script excerpts, images, and voices. For each one, note the source, allowed use, and restrictions. If the origin of a critical asset is unknown, exclude it from the pilot until the conditions are clarified.
3. Assign the creative decision-maker
The director, art director, or artist sets the selection criteria, chooses the option, and specifies what needs to be redone. Save versions and decisions: this helps the team understand the real contribution of the tool and the people, and it also helps resolve disputes after release.
4. Calculate the full cycle
Record the time for preparing inputs, generation, selection, refinement, approval, and rights review. Include compute costs and repeat generations. Comparison only makes sense for a comparable task with the same quality requirements.
5. Decide where the technology is appropriate
If the pilot delivers more workable options without losing creative control and with clear rights, expand usage gradually. If the team spends the saved time fixing frame consistency or cannot confirm rights to the source materials, return to the task definition. A negative pilot result is useful too: it shows the tool's limits.
Frequently asked questions
Will AI replace animation artists?
Individual tasks can be automated, and the mix of jobs and demand for roles can change. That does not support a verified conclusion that an entire profession will be replaced or that a specific percentage of jobs will be cut. A studio should separately assess the impact on employment, retraining, and compensation for the people whose work is changed by the new process.
Can AI be used in commercial animation?
Technically, yes. But before release, you need to verify rights to the source materials, the terms of the service you use, consent from people if their likeness or voice is used, and the rights to the final asset under applicable law and contracts. The mere fact that something was generated does not replace this review. WIPO guidance.
Is it true that AI will reduce animation costs by 90%?
Katzenberg made such a forecast about production time and cost. There is no universal confirmation for projects of different types in the sources reviewed. Check the savings in your own full production workflow, including revisions, quality control, and legal expenses.
How is generative AI different from Pixar computer animation?
CAPS and later computer animation systems changed how frames are built and processed. A generative model additionally creates material based on training data, so production questions also include training data, consent, output similarity, and rights allocation. The history of CAPS at Pixar, U.S. Copyright Office overview.
Where should a small studio or content team start?
Start with one internal task where you can compare quality and labor input without needing to publish the result right away. Record the data sources, who is responsible for the creative decision, and the completion criteria. After that, decide whether to expand the process.
How AgentSunrise can help with a pilot
For a creative team, AgentSunrise can help structure a limited experiment with generative production: break down the current material-preparation workflow, prepare a working prototype for image or video variations and set up review and handoff of the result to a person for approval. These actions align with the stated areas of generative production, process auditing, MVPs, and testing. AgentSunrise focus areas.
The first step is to choose one process, document its current labor input, material sources, rights restrictions, and acceptance criteria. After that, you can discuss the scope of the pilot and the team roles. Discuss the project.
Conclusion
The technology expands the set of available techniques, but it does not by itself distribute benefits and risks fairly. The lesson from the history of animation and sound recording is that people need to agree on the terms of labor use while the production process is changing. For studios, the practical path is a narrow pilot with measurable economics, clear rights to materials, and a creative decision made by a human.