Monolith Soft insists: it wasn’t generative AI, but good old procedural asset generation that saved them thousands of hours while building Xenoblade Chronicles 3’s vast world.
Three Monolith Soft developers—Yoichi Akizuki (map model designer), Mitsuhiro Hirose (support engineer), and Takashi Shibahara (programmer)—told CG World Japan that using procedural generation made it dramatically easier to create Xenoblade Chronicles 3’s enormous game world. The team gradually worked these tools into their workflow, especially with software like Houdini, allowing them to establish basic 3D models, textures, and rule sets to be used as input for asset-generating algorithms.
For example, a developer could create the 3D model and textures for a tree, define parameters for its appearance, and the software would then generate countless trees, each unique but all conforming to the original guidelines. Because procedural asset generation builds off assets and rules made by the team (rather than scraping data from the internet or other sources like generative AI), it avoids copyright controversies and keeps full creative control in-house.
Praised for its massive open world, Xenoblade Chronicles 3 owes a lot to this process. With automation handling the repetitive, less important asset creation, the development team had more time to focus on gameplay and core art. Akizuki noted that each new Xenoblade game required more and more unique assets, eventually making it impossible to keep up by hand.
“We brought in procedural generation to deal with the sheer increase in required assets. In previous games, we could manually create 1,000–2,000 items, but now we’re talking about 100,000. Costs and scheduling just made the old ways impossible. By automating and using procedural placement, we could refocus on actually building the game. We’d experimented with Houdini for a while, but Xenoblade Chronicles 3 was the first time we used it as part of our production pipeline,” Akizuki explained.
Asked which parts of development benefited most, Akizuki highlighted not just asset creation but collision detection: “It massively reduced the workload for placing fine details, and we also used it for core collision systems, which is a big deal in games. Collision is always the last crucial step, so we automated 70% of the process and left the final 30% to manual tweaks. That cut down dev time dramatically, freeing up resources to polish gameplay and visuals. Ultimately, Houdini made the game better by letting us focus where it matters most,” he added.
Shibahara pointed out that Houdini enables technical artists and designers to make tweaks directly, without having to tie up programmers for every adjustment: “Previously, automatic asset generation would bog down programmers and designers for weeks or months. Now, technical artists and designers can make changes directly, so programmers can focus on improving the environment — everyone wins,” said Shibahara.
Hirose said Monolith’s R&D team is now exploring how far automation can go, including procedural cityscapes: “Designers just place grey boxes, and with a single click, the system builds complete buildings. It assembles houses from modules, calculates the right number of floors, arranges roofs, and outputs the result in a game-ready format,” he explained.
The studio was upfront about “cutting corners” with automation, but stressed that these lessons were learned on a finished game, not a work-in-progress.
Leave a Reply