Optimizing Digital Double Creation and Animation Workflows

This project addressed the challenge of achieving high-fidelity automatic facial animation, particularly noting limitations in “out-of-the-box” mouth and tongue movements. The research explored and compared digital double creation and animation workflows using Unreal’s Metahuman Creator and Reallusion’s Character Creator.

The methodology involved rigorous testing of motion capture retargeting, evaluating visual and performance accuracy, and developing three refined production pipelines for photogrammetry cleanup, Metahuman creation, and Character Creator integration. We successfully produced eight distinct prototypes, demonstrating enhanced realism through custom techniques like Wrap and Substance Painter.

This foundational work provides a clear understanding of system gaps, refined motion capture processes, and reusable documentation, significantly enhancing character accuracy and animation quality for cross-platform applications.

Facial Animation Accuracy Requires Smarter Debugging Tools

Automatic facial animation often struggles with mouth and eye inaccuracies, particularly in real-time applications. For digital doubles used in virtual production or game engines, these flaws undermine realism and cross-platform compatibility. To address this, the project sought to build a visual debugging workflow that supports accurate animation transfer and evaluation across different rigging systems—such as MetaHuman, Character Creator, and Unity. The goal was to compare and refine these workflows, improving fidelity from raw motion capture to final render. By identifying these system gaps, the research aimed to lay the groundwork for refining character accuracy and expanding performance capture capabilities across various platforms.

Designing a Modular Workflow to Improve Cross-Platform Facial Animation

Eight prototypes were developed by creating two to three asset variations for each of three scanned actors: a raw photogrammetry scan, a standard Metahuman version, and a Wrapped Metahuman with baked high-fidelity textures. These prototypes facilitated the comparison of visual accuracy, animation performance, and cross-platform compatibility, providing a crucial basis for refining future digital double workflows.

The entire development process, tool usage, and findings were meticulously documented. This documentation supports replication, future development, and broader adoption of improved animation debugging practices.

Enabling Consistent and Realistic Facial Animation

The project delivered eight animation-ready prototypes and three cross-rig production workflows, enhanced the understanding and capabilities in facial animation, specifically:

• Improved rig compatibility and animation fidelity via custom mesh and texture workflows.
• Granular control of facial expression correction using MetaHuman Animator and Maya-based blendshapes.

This work lays a foundation for scalable, platform-agnostic facial animation debugging tools—supporting not only more believable digital humans, but also accelerating workflows in film, games, and immersive media.