Adobe and Disney Build Custom Generative AI Models for Production

Adobe and The Walt Disney Company have announced a multi-year strategic alliance under which Adobe will give Disney early access to its latest technologies for use across Disney's theme parks, cruise line and ESPN content. The alliance was published in Adobe's newsroom on November 9, 2026 (Adobe newsroom).

The work is built on Adobe Firefly Foundry, which Adobe announced at Adobe MAX in October 2025. Firefly Foundry lets businesses work directly with Adobe to create tailored generative AI, meaning models built for a specific client rather than rented as a general-purpose tool. Under the Disney arrangement, the two companies are developing proprietary, artist-led generative AI models to support Disney's production pipeline. According to The Verge, those customized models are trained on Disney properties' world knowledge, including character attributes, environments and visual styles, with the stated aim of producing commercially safe, production-ready outputs.

The models are described as human-centered, and Disney will work directly with Adobe on them. The difference lies in the training data. Rather than licensing a general image or video model and prompting it toward a franchise, Disney is commissioning models whose training data is the franchise itself, with artists in the loop. Whether that produces measurable gains in throughput or cost is not something the announcement quantifies.

The Verge also reported that Disney's Imagineering R&D division already uses Adobe's Firefly AI platform for theme park development. The alliance therefore formalizes an existing working relationship in Imagineering and extends the scope to other parts of the company, which the reporting identifies as the theme parks, the cruise line and ESPN.

The scope is broad. A theme park attraction, a cruise ship interior and a sports broadcast graphics package have different production cycles, asset requirements and approval chains. Applying one custom model family across all three implies a shared asset pipeline and a common set of rights-managed training inputs. The announcements do not describe the technical architecture, the hosting arrangements or how outputs will be reviewed before release.

The broader context here is the shift in how large rights holders are engaging with generative AI. The early phase was largely defined by general models trained on broad web data, with studios and brands working around the legal and brand-safety questions those models raised. A custom model trained on owned intellectual property, supplied by a vendor that already sells enterprise creative tooling, addresses those questions at the procurement level rather than the prompt level. That moves the governance problem into the contract and the training data, where rights holders have more control.

In my view, the more interesting question for the industry is less whether studios will use generative models than who will own the models that encode their characters and worlds. This announcement answers that for Disney in a specific way, and other rights holders will be watching the terms.

For Adobe, the alliance gives Firefly Foundry a flagship enterprise reference of the kind a custom-model business needs. For Disney, the stated benefit is faster access to Adobe's latest tools in its production work. Both sides are describing intent and forward arrangements.

If the models perform as described, the result would be generative tooling steeped in the studio's own visual language and shaped by its artists. That is a more constrained proposition than the generic image generator most people first encountered, and it is the direction the alliance points.