Improved developer efficiency
We’re continuously looking at ways to improve efficiency for developers, so they can focus on the creative process. Fortnite’s adoption of new graphics technologies such as Nanite that have a larger disk footprint challenged us to find ways to reduce the time our teams spend loading the game or just getting content from Perforce.
The large and very distributed Fortnite team benefited greatly from new Unreal Cloud DDC, a globally efficient model for our derived data cache that provides fast access to optimized versions of content such as textures or meshes, avoiding the need for local processing. As a result, Editor load times on Fortnite are at least two times faster.
Virtual Assets have reduced disk footprint for content and consequently sync times for Fortnite by over four times for a single workspace; even more savings can be gained through de-duplication of any data common to multiple workspaces. We strip the large bulk data such as raw texture pixels from the Unreal Asset’s much smaller class properties such as texture resolution, format, and so on, and store them separately. From the outside, Virtual Assets are indistinguishable from regular Unreal Assets but they are many times smaller. Developer workflows remain largely unchanged while clients spend significantly less time waiting for their Perforce syncs, only downloading data they actually need to use in the editor as they need it.
Unreal Engine’s Machine Learning (ML) Deformer enables you to create high-fidelity approximations of complex proprietary rigs—or any arbitrary deformation—by using a custom Maya plugin to train a machine learning model, which in turn runs in real time in Unreal Engine.
With Chapter 4, the Fortnite team used the ML Deformer to allow for higher-quality character deformations on muscles and cloth, with the Hulk and Sunlit characters being the first to benefit. The team was able to capture complex deformation and simulation from external DCCs and inject that geometric result into their gameplay rigs. As a result of this battle-testing, the ML Deformer moved from Experimental to Beta in UE 5.1, and the team’s close collaboration with the ML Deformer engineers offers the opportunity for the technology to be pushed further forward with new features and stability fixes planned for UE 5.2.
In Chapter 4, the Fortnite team worked closely with developers from 3Lateral and the MetaHuman team to upgrade the facial animation capabilities of Fortnite characters. This included several updates to character quality, such as increasing the number of expressions and joints on new characters, as well as creating new rigging and animation tools to push facial animation to new heights.
As a result, Unreal Engine 5.1 offers improvements to the runtime functionality of Rig Logic to adhere to performance constraints, as well as remapping nodes created specifically to support forwards and backwards compatibility of MetaHuman technology in Fortnite.
We hope you enjoy all of the improvements we’ve made to Unreal Engine 5.1 during development of Fortnite Battle Royal Chapter 4, and find them useful in taking your own titles to the next level of visual fidelity. We look forward to seeing the results!
Credit: Source link