← ArchivePaper2025
One Model to Rig Them All: Diverse Skeleton Rigging with UniRig
Jia-Peng Zhang, Cheng-Feng Pu, Meng-Hao Guo, Yan-Pei Cao, Shi-Min Hu
Large autoregressive transformer generates skeletons and skinning weights for diverse 3D asset categories in a single unified framework.
Abstract
The rapid evolution of 3D content creation, encompassing both AI-powered methods and traditional workflows, is driving an unprecedented demand for automated rigging solutions that can keep pace with the increasing complexity and diversity of 3D models. We introduce UniRig, a novel, unified framework for automatic skeletal rigging that leverages the power of large autoregressive models and a bone-point cross-attention mechanism to generate both high-quality skeletons and skinning weights. Unlike previous methods that struggle with complex or non-standard topologies, UniRig accurately predicts topologically valid skeleton structures thanks to a new Skeleton Tree Tokenization method that efficiently encodes hierarchical relationships within the skeleton. To train and evaluate UniRig, we present Rig-XL, a new large-scale dataset of over 14,000 rigged 3D models spanning a wide range of categories. UniRig significantly outperforms state-of-the-art academic and commercial methods, achieving a 215% improvement in rigging accuracy and a 194% improvement in motion accuracy on challenging datasets. Our method works seamlessly across diverse object categories, from detailed anime characters to complex organic and inorganic structures, demonstrating its versatility and robustness.
How to read this
- Category
- Method: a unified autoregressive auto-rigging framework
- Contributions
- UniRig, a single autoregressive-transformer framework generating both skeletons and skinning weights via a bone-point cross-attention mechanism
- A Skeleton Tree Tokenization scheme that encodes hierarchy and yields topologically valid skeletons across diverse, non-standard topologies
- Rig-XL, a large-scale dataset of over 14,000 rigged 3D models spanning many categories
- Context
- Extends learning-based auto-rigging beyond RigNet (Xu et al. 2020), recasting skeleton plus skinning prediction as autoregressive sequence generation in the style of large transformer models.Builds on: RigNet: Neural Rigging for Articulated Characters
- Correctness
- Reported gains in rigging and motion accuracy are measured against academic and commercial baselines on challenging datasets, but results lean on the new Rig-XL training set, so generalization to categories or topologies under-represented there should be checked.
- Clarity
- Likely accessible at the idea level; a first pass conveys the tokenization-plus-cross-attention story, with a second pass needed for the tokenizer and training formulation.
- How to read it
- First pass on the abstract and the Skeleton Tree Tokenization figure to grasp how hierarchy becomes a token sequence; do a second pass on the bone-point cross-attention and Rig-XL composition if you care about reproducibility or coverage.
Builds on
Built upon by
Nothing yet.
Related work
- S3: Neural Shape, Skeleton, and Skinning Fields for 3D Human Modeling 2021 / CVPR
- ASMR: Adaptive Skeleton-Mesh Rigging and Skinning via 2D Generative Prior 2025 / CGF
- MoRig: Motion-Aware Rigging of Character Meshes from Point Clouds 2022 / SIGGRAPH Asia
- NeuroSkinning: Automatic Skin Binding for Production Characters with Deep Graph Networks 2019 / SIGGRAPH
Keywords
This page summarises the entry and links to its original source. The archive never hosts or redistributes the publication itself.Show it in the full archive list →