Adobe, Disney build AI design models for theme park work

Adobe and Walt Disney Imagineering have developed a custom artificial intelligence system for theme park design using Adobe Firefly Foundry and Disney's own intellectual property.

The collaboration centres on bespoke generative AI models trained on Imagineering's design catalogue rather than broad internet data sets. The tools are designed to work within the creative software and workflows already used by Disney's design teams.

Walt Disney Imagineering, the division behind Disney's parks and attractions, plans to use the models across early design and pre-production. Initial applications include turning rough sketches into rendered 2D concept art, generating franchise-specific images and converting 2D designs into 3D prototypes for planning and engineering coordination.

The image model was developed on Imagineering assets and can produce material aligned with franchises including Mickey & Friends, Frozen, Moana, Lilo & Stitch and Cars. A separate sketch-to-image model is intended to help creative teams move from hand-drawn concepts to more finished visualisations earlier in the process.

Another part of the system is a 3D modelling process that turns 2D renderings and concepts into prototypes. Disney said this could help teams plan builds, estimate materials and work with engineering groups before construction begins.

The move reflects a broader push by media and entertainment companies to use generative AI in internal production while limiting the legal and brand risks associated with open models trained on scraped web content. For Disney, whose business depends heavily on tightly controlled characters and visual worlds, provenance and consistency are especially important.

Adobe has been expanding its efforts to offer custom AI models for corporate customers through Firefly Foundry. The offering lets organisations work with Adobe on models tailored to their own brand assets and internal requirements, rather than relying only on general-purpose image generators.

In Disney's case, the model is positioned as a way to preserve visual accuracy while reducing the time needed to move from concept to prototype. The project also shows how large entertainment companies are trying to keep tighter control over the data used to train creative tools.

Creative workflows

The new workflows are aimed at the design and pre-visualisation stages rather than finished guest-facing outputs. That places the tools in one of the more commercially practical areas of AI adoption, where businesses can use automation to speed repetitive or exploratory work while retaining human oversight over final creative decisions.

Adobe described the partnership as an example of how custom models can be applied where intellectual property is central to a product's value. Disney described the work as part of a longstanding approach that combines technology with creative development.

"Storytelling is in Disney's DNA. Empowering creators with the latest AI innovations is in ours," said Hannah Elsakr, Vice President, GenAI New Business Ventures at Adobe.

"As the teams at Imagineering build new experiences for fans around the world, our tools and workflows will provide a creative foundation to explore bolder ideas and make the best ones a reality."

Disney emphasised speed and consistency in applying the technology to park development. The company is expanding across parks, hotels and cruise operations, increasing the volume of design work tied to its film and television franchises.

"At Imagineering, we've always believed technology and human creativity can work together responsibly," said Kyle Laughlin, Senior Vice President of Research and Development, Technology and Engineering at Walt Disney Imagineering.

"Our work with Adobe lets us bring Disney stories and characters to life in our parks faster, and with the emotional quality our guests expect."

Brand control

The collaboration highlights one of the central tensions in commercial AI deployment: companies want the efficiency gains of generative systems without exposing themselves to uncertain training data, brand drift or potential ownership disputes. By training on licensed and proprietary assets, Adobe and Disney are presenting this project as a controlled alternative for a rights-sensitive industry.

That approach may prove relevant beyond theme parks. Film studios, broadcasters, games companies and consumer brands face similar concerns when using AI to create material that must stay closely aligned with established characters, designs and visual identities.

For Disney Imagineering, the immediate application is more concrete: a design toolkit intended to turn sketches and reference material into concept art, franchise-specific imagery and 3D prototypes within the same development pipeline used to shape attractions and other physical experiences.