Scaling a Virtual Gallery with USD and Unreal Engine

Designing a virtual art gallery for a recent short film, I aimed for a massive, adaptable space reflective of an affluent collector. Our previous attempts at building large-scale, intricate environments in Unreal directly were often bogged down by performance hits and complex asset management. This time, we decided to lean heavily into USD (Universal Scene Description) workflows.

What worked incredibly well was building modular wall sections, art plinths, and even lighting rigs in Maya and then importing them as individual USD assets. We then assembled these within Unreal using a dedicated USD stage, instancing elements repeatedly. This allowed for incredibly fast iterations on the layout and a significant performance boost. When a director asked for a ceiling design change, we could update a single USD asset in Maya, and it would propagate across all instances in Unreal almost instantly. The scalability was fantastic.

What didn't work as smoothly was initially trying to integrate complex, animated character rigs also as USD. While theoretically possible, the animation pipeline between Maya and Unreal via USD, for complex rigs with blend shapes and physics, still felt a bit cumbersome and less direct than our established FBX pipeline for characters. It required more troubleshooting than we had time for. Overall, it pushed us to think about environment architecture in a more distributed, flexible way. Do others find USD for environments to be a game-changer, but still prefer FBX for complex character animation, or am I missing a trick?