Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Li, Zhaoyang, Qian, Dongjun, Su, Kai, Diao, Qishuai, Xia, Xiangyang, Liu, Chang, Yang, Wenfei, Zhang, Tianzhu, Yuan, Zehuan
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.00438
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866914363703558144
author Li, Zhaoyang
Qian, Dongjun
Su, Kai
Diao, Qishuai
Xia, Xiangyang
Liu, Chang
Yang, Wenfei
Zhang, Tianzhu
Yuan, Zehuan
author_facet Li, Zhaoyang
Qian, Dongjun
Su, Kai
Diao, Qishuai
Xia, Xiangyang
Liu, Chang
Yang, Wenfei
Zhang, Tianzhu
Yuan, Zehuan
contents Diffusion Transformer has shown remarkable abilities in generating high-fidelity videos, delivering visually coherent frames and rich details over extended durations. However, existing video generation models still fall short in subject-consistent video generation due to an inherent difficulty in parsing prompts that specify complex spatial relationships, temporal logic, and interactions among multiple subjects. To address this issue, we propose BindWeave, a unified framework that handles a broad range of subject-to-video scenarios from single-subject cases to complex multi-subject scenes with heterogeneous entities. To bind complex prompt semantics to concrete visual subjects, we introduce an MLLM-DiT framework in which a pretrained multimodal large language model performs deep cross-modal reasoning to ground entities and disentangle roles, attributes, and interactions, yielding subject-aware hidden states that condition the diffusion transformer for high-fidelity subject-consistent video generation. Experiments on the OpenS2V benchmark demonstrate that our method achieves superior performance across subject consistency, naturalness, and text relevance in generated videos, outperforming existing open-source and commercial models.
format Preprint
id arxiv_https___arxiv_org_abs_2510_00438
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BindWeave: Subject-Consistent Video Generation via Cross-Modal Integration
Li, Zhaoyang
Qian, Dongjun
Su, Kai
Diao, Qishuai
Xia, Xiangyang
Liu, Chang
Yang, Wenfei
Zhang, Tianzhu
Yuan, Zehuan
Computer Vision and Pattern Recognition
Diffusion Transformer has shown remarkable abilities in generating high-fidelity videos, delivering visually coherent frames and rich details over extended durations. However, existing video generation models still fall short in subject-consistent video generation due to an inherent difficulty in parsing prompts that specify complex spatial relationships, temporal logic, and interactions among multiple subjects. To address this issue, we propose BindWeave, a unified framework that handles a broad range of subject-to-video scenarios from single-subject cases to complex multi-subject scenes with heterogeneous entities. To bind complex prompt semantics to concrete visual subjects, we introduce an MLLM-DiT framework in which a pretrained multimodal large language model performs deep cross-modal reasoning to ground entities and disentangle roles, attributes, and interactions, yielding subject-aware hidden states that condition the diffusion transformer for high-fidelity subject-consistent video generation. Experiments on the OpenS2V benchmark demonstrate that our method achieves superior performance across subject consistency, naturalness, and text relevance in generated videos, outperforming existing open-source and commercial models.
title BindWeave: Subject-Consistent Video Generation via Cross-Modal Integration
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2510.00438