AI & ML New Capability

SOMA provides a unified, differentiable layer that bridges incompatible human body models like SMPL and SMPL-X in a single closed-form pass.

arXiv · March 18, 2026 · 2603.16858

Jun Saito, Jiefeng Li, Michael de Ruyter, Miguel Guerrero, Edy Lim, Ehsan Hassani, Roger Blanco Ribera, Hyejin Moon, Magdalena Dadela, Marco Di Lucca, Qiao Wang, Xueting Li, Jan Kautz, Simon Yuen, Umar Iqbal

The Takeaway

It eliminates the need for complex, per-model retargeting or adapters in human reconstruction pipelines. This allows researchers to mix diverse identity sources and motion datasets seamlessly, significantly simplifying the workflow for 3D animation and simulation.

From the abstract

Parametric human body models are foundational to human reconstruction, animation, and simulation, yet they remain mutually incompatible: SMPL, SMPL-X, MHR, Anny, and related models each diverge in mesh topology, skeletal structure, shape parameterization, and unit convention, making it impractical to exploit their complementary strengths within a single pipeline. We present SOMA, a unified body layer that bridges these heterogeneous representations through three abstraction layers. Mesh topology