Zijie Wu

Zijie Wu (Jarrent Wu)

Ph.D. Student in Computer Vision
Huazhong University of Science and Technology (HUST)

I am a third-year Ph.D. student supervised by Prof. Xiang Bai. My research interests include Image/Video/3D/4D Generation, and Motion Synthesis. I am nearing graduation and will soon be on the job market. Should my research or background be of interest to you, please do not hesitate to contact me via email: zjwu1031@gmail.com

💼 Experience

Research Intern (Qingyun Plan)

ARC Lab, Tencent (腾讯ARC Lab) May 2026 - Present
  • Selected for the Tencent Qingyun Plan (腾讯青云计划), a top-tier talent program.
  • Focusing on World Model & Long Video Gen.

Research Intern (Qingyun Plan)

Hunyuan3D, Tencent (腾讯混元3D) Aug 2025 - Apr 2026
  • Selected for the Tencent Qingyun Plan (腾讯青云计划), a top-tier talent program.
  • Focusing on Mesh Animation and Motion Synthesis.
  • Published R-DMesh at Siggraph2026 as the first author.

Research Intern

DAMO Academy, Alibaba Group (阿里巴巴达摩院) May 2023 - Jul 2025
  • Conducted research on 2D/3D/4D Generation.
  • Published SC4D and AnimateAnyMesh at ECCV2024 and ICCV2025 as the first author.

📝 Selected Publications

* denotes equal contribution.

Paper Image

R-DMesh: Video-Guided Mesh Animation via Rectified Dynamic Mesh Flow

Zijie Wu, Lixin Xu, Puhua Jiang, Sicong Liu, Chunchao Guo, Xiang Bai
Siggraph, 2026
A novel feed-forward video-guided mesh animation framework that rectifies the initial pose of the input mesh according to the first frame of the guidance video, which enables downstream applications like: pose/motion retargeting, holistic 4D generation.
[Paper] [Project Page] [Code] [Video]
Paper Image

AnimateAnyMesh++: A Flexible 4D Foundation Model for High-Fidelity Text-Driven Mesh Animation

Zijie Wu, Chaohui Yu, Fan Wang, Xiang Bai
TPAMI, Submitted
Proposed AnimateAnyMesh++, a flexible 4D foundation model for high-fidelity text-driven mesh animation. Designed DyMeshVAE-Flex, PLTA Attention, and Vertex Normal Injection to enable variable-length generation and superior local geometry preservation. Empowered by the introduced large-scale DyMesh-XL dataset, it surpasses existing methods in performance and efficiency, significantly broadening the prospects of 4D generation.
[Paper] [Code]
Paper Image

AnimateAnyMesh: A Feed-Forward 4D Foundation Model for Text-Driven Universal Mesh Animation

Zijie Wu, Chaohui Yu, Fan Wang, Xiang Bai
ICCV, 2025
Proposed AnimateAnyMesh, the first feed-forward universal mesh animation framework. Designed DyMeshVAE and Text-to-Trajectory RF Model to compress and generate vertex trajectories, eliminating the dependency on rigging and skinning and pioneering the new paradigm: feed-forward vertex-based 3D animation.
[Paper] [Project Page] [Code] [Video]
Paper Image

SC4D: Sparse-Controlled Video-to-4D Generation and Motion Transfer

Zijie Wu, Chaohui Yu, Yanqin Jiang, Chenjie Cao, Fan Wang, Xiang Bai
ECCV, 2024
We propopse SC4D, which disentangles motion and appearance synthesis into sparse point control and dense gaussians to achieve superior performance. Benifit from this disentangled representation, SC4D enables motion transfer applications.
[Paper] [Project Page] [Code] [Video]
Paper Image

CCPL: Contrastive Coherence Preserving Loss for Versatile Style Transfer

Zijie Wu*, Zhen Zhu*, Junping Du, Xiang Bai
ECCV, 2022 Oral Presentation
A generic contrastive loss is proposed to alleviate the flickering artifacts in video style transfer and improve the generation quality of image style transfer.
[Paper] [Code] [Web Demo] [Video]