Saved in:
Bibliographic Details
Main Authors: Xie, Zhenyu, Xia, Ji, Kampffmeyer, Michael, Hu, Panwen, Ma, Zehua, Zheng, Yujian, Wang, Jing, Chong, Zheng, Zhang, Xujie, Cheng, Xianhang, Liang, Xiaodan, Li, Hao
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2603.15415
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908889654493184
author Xie, Zhenyu
Xia, Ji
Kampffmeyer, Michael
Hu, Panwen
Ma, Zehua
Zheng, Yujian
Wang, Jing
Chong, Zheng
Zhang, Xujie
Cheng, Xianhang
Liang, Xiaodan
Li, Hao
author_facet Xie, Zhenyu
Xia, Ji
Kampffmeyer, Michael
Hu, Panwen
Ma, Zehua
Zheng, Yujian
Wang, Jing
Chong, Zheng
Zhang, Xujie
Cheng, Xianhang
Liang, Xiaodan
Li, Hao
contents Controllable character animation has advanced rapidly in recent years, yet multi-character animation remains underexplored. As the number of characters grows, multi-character reference encoding becomes more susceptible to latent identity entanglement, resulting in identity bleeding and reduced controllability. Moreover, learning precise and spatio-temporally consistent correspondences between reference identities and driving pose sequences becomes increasingly challenging, often leading to identity-pose mis-binding and inconsistency in generated videos. To address these challenges, we propose AnyCrowd, a Diffusion Transformer (DiT)-based video generation framework capable of scaling to an arbitrary number of characters. Specifically, we first introduce an Instance-Isolated Latent Representation (IILR), which encodes character instances independently prior to DiT processing to prevent latent identity entanglement. Building on this disentangled representation, we further propose Tri-Stage Decoupled Attention (TSDA) to bind identities to driving poses by decomposing self-attention into: (i) instance-aware foreground attention, (ii) background-centric interaction, and (iii) global foreground-background coordination. Furthermore, to mitigate token ambiguity in overlapping regions, an Adaptive Gated Fusion (AGF) module is integrated within TSDA to predict identity-aware weights, effectively fusing competing token groups into identity-consistent representations...
format Preprint
id arxiv_https___arxiv_org_abs_2603_15415
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle AnyCrowd: Instance-Isolated Identity-Pose Binding for Arbitrary Multi-Character Animation
Xie, Zhenyu
Xia, Ji
Kampffmeyer, Michael
Hu, Panwen
Ma, Zehua
Zheng, Yujian
Wang, Jing
Chong, Zheng
Zhang, Xujie
Cheng, Xianhang
Liang, Xiaodan
Li, Hao
Computer Vision and Pattern Recognition
Controllable character animation has advanced rapidly in recent years, yet multi-character animation remains underexplored. As the number of characters grows, multi-character reference encoding becomes more susceptible to latent identity entanglement, resulting in identity bleeding and reduced controllability. Moreover, learning precise and spatio-temporally consistent correspondences between reference identities and driving pose sequences becomes increasingly challenging, often leading to identity-pose mis-binding and inconsistency in generated videos. To address these challenges, we propose AnyCrowd, a Diffusion Transformer (DiT)-based video generation framework capable of scaling to an arbitrary number of characters. Specifically, we first introduce an Instance-Isolated Latent Representation (IILR), which encodes character instances independently prior to DiT processing to prevent latent identity entanglement. Building on this disentangled representation, we further propose Tri-Stage Decoupled Attention (TSDA) to bind identities to driving poses by decomposing self-attention into: (i) instance-aware foreground attention, (ii) background-centric interaction, and (iii) global foreground-background coordination. Furthermore, to mitigate token ambiguity in overlapping regions, an Adaptive Gated Fusion (AGF) module is integrated within TSDA to predict identity-aware weights, effectively fusing competing token groups into identity-consistent representations...
title AnyCrowd: Instance-Isolated Identity-Pose Binding for Arbitrary Multi-Character Animation
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.15415