Disney turns to Adobe AI to speed park design

Disney’s bespoke – Walt Disney Imagineering is using a bespoke AI model built on Adobe’s Firefly Foundry to generate Disney-branded attraction and environmental design faster—from sketch-to-image and character assets to 3D prototypes—while keeping “human in the loop” workflows.
A Disney concept sketch now has a new shortcut to reality—fast enough that Imagineers may be able to catch design mistakes before anything is built in the field.
Walt Disney Imagineering. the design and engineering arm behind Disney parks and cruises. has developed its own bespoke AI model from Adobe. The goal is to use the technology to create attractions and environmental design—such as storefronts and architecture—at a pace “that wouldn’t otherwise be possible. ” with outputs that remain aligned to Disney’s brand guide.
The custom AI is built on Firefly Foundry, Adobe’s boutique AI service that trains on a brand’s IP and catalog. For Imagineering, that required ingesting decades of internal work, including artist drawings plus “every architectural diagram and piece of concept art they’ve ever created.”
“We’re able to unify for the very first time” multiple systems that are currently scattered across the company. Walt Disney Imagineering senior vice president of R&D technology and engineering Kyle Laughlin said. He described the AI’s reach in physical terms—“It’s on people’s laptops. it’s underneath people’s desks. it is in people’s brains”—and added that it “literally is in dozens of disparate systems.”.
Technically, the model is designed with billions of parameters and is capable of generating on-brand Disney assets. It includes a sketch-to-image model that can turn hand-drawn concepts into 2D concept art. There is also a custom image model built for franchise-accurate creative assets for characters such as Mickey Mouse and Lilo and Stitch. For visualization, the 3D-modeling tool can convert 2D renderings into prototypes.
For Imagineering. the payoff isn’t just speed—it’s the ability to move from early design ideas to more realistic previews quickly. Laughlin said project timelines that normally last “five, six, seven years” can be compressed. He also said iteration work that once took “months” will take “days,” which changes how the team can handle adjustments.
Instead of waiting until construction is underway, Imagineers will be able to pre-visualize assets of things like building or restaurant facades in a reality headset, making changes early—when tweaks are less expensive than in-field corrections.
That matters because Disney’s pressure to deliver is tied to a massive commitment it made in 2023. After the Walt Disney Company pledged $60 billion in investments over the next decade in its experiences business. Imagineering is expected to translate that ambition into new and upgraded theme-park. cruise. hotel. and food-and-beverage experiences around the world.
Laughlin framed that undertaking as unusually large for the unit: “This is an unprecedented amount of capital going into everything from theme parks. cruise ships. hotels. food and beverage experiences around the world. and the ability to deliver against that ambition is unlike anything we’ve seen in the history of Imagineering.”.
Still, keeping every output on brand is a real constraint, not a branding slogan. Hannah Elsakr. Adobe’s head of new GenAI business ventures. said an iconic brand is not a single file stored in a library. “It’s not a flat file sitting in an asset library. it’s not just a font that we recognize. ” she said. “It’s really how the thousands of small decisions that the creative leaders make each day around creative intent.”.
The rollout arrives amid a wider conversation about AI in entertainment. Late last year. Disney announced an agreement with OpenAI to license its characters to Sora. the company’s short-lived AI short-form video app that shut down in April. After Disney laid off about 1,000 employees earlier this year, some Marvel fans blamed AI.
Inside Imagineering’s workflow, Laughlin said the key guardrail is people. “The most important thing about our workflows is that there are human in the loop,” he said. He also pointed to a practical limit on staffing: Imagineering “simply can’t find enough resources to do all of the work we have in front of us.”.
For that reason, he said the ability to compress timelines using tools and workflows like the AI model is “incredibly important” to deliver the $60 billion in experiences investment.
Disney’s broader business context adds to the urgency. The company’s parks business, along with streaming, has helped it grow to about $9.5 billion in revenue, up 7% year-over-year.
Taken together, the message from Disney Imagineering is clear: the technology isn’t being introduced to replace the creative process, but to compress the time between early concept and build-ready design—before the most expensive decisions are made in the real world.
Disney Walt Disney Imagineering Adobe Firefly Foundry generative AI theme parks park design sketch-to-image 3D modeling human in the loop OpenAI Sora Kyle Laughlin Hannah Elsakr experiences investment
So Disney is basically making park designs with AI now? Kinda wild but also… who’s checking the details if it’s just gonna spit out “on brand” stuff.
I read that it catches design mistakes before they build in the field and that’s honestly the only part I care about. Also Firefly Foundry?? Isn’t that like Adobe stuff for school kids lol.
Wait so this is the same as when they “AI generate” characters and it ends up looking kinda off? Like if it’s trained on all their internal drawings wouldn’t it just copy the old stuff and stop being creative. Or maybe they’ll just print whatever Disney wants with a laptop…
This headline makes it sound like they’re using AI to build entire attractions overnight which is scary in a way. But then it says human in the loop so I guess humans still save it. I just don’t get why they need Adobe for it, like couldn’t they do their own thing? My cousin said Adobe AI is “under desks in people’s brains” or something, so now I’m confused.