AniGen: Unified S3 Fields for Animatable 3D Asset Generation

Yi-Hua Huang1 Zi-Xin Zou2 Yuting He3 Chirui Chang1 Cheng-Feng Pu4 Ziyi Yang1 Yuan-Chen Guo2 Yan-Pei Cao2† Xiaojuan Qi1†
1The University of Hong Kong 2VAST 3The Chinese University of Hong Kong 4Tsinghua University
† Corresponding authors
Paper ArXiv Code 🤗 Demo Tripo Video

Given a single conditional image, AniGen generates a 3D shape along with its skeleton and skinning weights, supporting animals, humanoids, and machinery.

Abstract

Animatable 3D assets, defined as geometry equipped with an articulated skeleton and skinning weights, are fundamental to interactive graphics, embodied agents, and animation production. While recent 3D generative models can synthesize visually plausible shapes from images, the results are typically static. Obtaining usable rigs via post-hoc auto-rigging is brittle and often produces skeletons that are topologically inconsistent with the generated geometry. We present AniGen, a unified framework that directly generates animate-ready 3D assets conditioned on a single image. Our key insight is to represent shape, skeleton, and skinning as mutually consistent S3 Fields (Shape, Skeleton, Skin) defined over a shared spatial domain. To enable the robust learning of these fields, we introduce two technical innovations: (i) a confidence-decaying skeleton field that explicitly handles the geometric ambiguity of bone prediction at Voronoi boundaries, and (ii) a dual skin feature field that decouples skinning weights from specific joint counts, allowing a fixed-architecture network to predict rigs of arbitrary complexity. Built upon a two-stage flow-matching pipeline, AniGen first synthesizes a sparse structural scaffold and then generates dense geometry and articulation in a structured latent space. Extensive experiments demonstrate that AniGen substantially outperforms state-of-the-art sequential baselines in rig validity and animation quality, generalizing effectively to in-the-wild images across diverse categories including animals, humanoids, and machinery.

Video

Results

Each row shows one subject across six viewpoints. The last three columns overlay the generated skeleton.

360° Rotation
Initial Pose 360°
Fixed Camera
360° + Skeleton
Initial Pose + Skeleton
Fixed Camera + Skeleton
Dog Dog
Eagle Eagle
Horse Horse
Bird Bird
Whale Whale
Iron Boy Iron Boy
Evo Evo
Mairo Mairo
Child Child
BrickBob BrickBob
Bruno Star Bruno Star
Woman Woman
Machine Arm Machine Arm
Machine Hand Machine Hand
Macbook Macbook
Plant Plant
Machine Dog Machine Dog
Money Tree Money Tree

Comparisons

AniGen vs. state-of-the-art auto-rigging baselines. Second row shows skeleton overlay.

Method

AniGen co-generates shape, skeleton, and skinning through unified S3 Fields. We introduce a confidence-decaying skeleton field to handle geometric ambiguity at bone prediction boundaries, and a dual skin feature field that enables joint-count agnostic skinning. A two-stage flow-matching pipeline first synthesizes sparse structure, then generates dense geometry and articulation.

AniGen pipeline overview
Pipeline overview. AniGen encodes shapes with their skeletons and skinning into unified S3 Fields, then compresses them into structured latents via auto-encoding. A two-stage flow model generates these structured latents from image-conditioned noise.
Two-stage flow model architecture
Two-stage flow model. Stage 1 (Sparse Structure Flow) generates a sparse structural scaffold. Stage 2 (Structured Latent Flow) produces dense geometry and articulation conditioned on the scaffold.

Citation

@article{huang2026anigen,
  title     = {AniGen: Unified $S^3$ Fields for Animatable 3D Asset Generation},
  author    = {Huang, Yi-Hua and Zhou, Zi-Xin and He, Yuting and Chang, Chirui
               and Pu, Cheng-Feng and Yang, Ziyi and Guo, Yuan-Chen
               and Cao, Yan-Pei and Qi, Xiaojuan},
  journal   = {ACM SIGGRAPH},
  year      = {2026}
}