Zhou Zixiang's Homepage

Zhou Zixiang

I am currently a researcher at Tencent HunYuan. My research interests include video generation and multimodal human-centric generation.

Before joining Hunyuan, I spent wonderful two and half years at xiaobing.ai, where I worked on multimodal human motion generation and understanding. Before that I was a research and development engineer at Baidu IDG, and a senior engineer at Huawei ADS, respectively. I received my Ph.D. degree from Rutgers, the State University of New Jersey, where I was fortunate to be supervised by Prof. Jie Gong in 2018. I worked as a research intern at Bentley System with Dr. Zheng Yi Wu in 2017.

zhouzixiang [at] tencent.com

Scholar Github

Publications

Euphonium: Steering Video Flow Matching via Process Reward Gradient Guided Stochastic Dynamics

Preprint

Ruizhe Zhong Jiesong Lian Xiaoyue Mi Zixiang Zhou Yuan Zhou Qinglin Lu Junchi Yan

Zhou Zixiang

Publications

Video Generation

Euphonium: Steering Video Flow Matching via Process Reward Gradient Guided Stochastic Dynamics

Making Avatars Interact: Towards Text-Driven Human-Object Interaction for Controllable Talking Avatars

StreamAvatar: Streaming Diffusion Models for Real-Time Interactive Human Avatars

ActAvatar: Temporally-Aware Precise Action Control for Talking Avatars

SoliReward: Mitigating Susceptibility to Reward Hacking and Annotation Noise in Video Generation Reward Models

Video Generation Models are Good Latent Reward Models

Hunyuanvideo 1.5 technical report

UniAVGen: Unified Audio and Video Generation with Asymmetric Cross-Modal Interactions

Hunyuanvideo-homa: Generic human-object interaction in multimodal driven human animation

Hunyuanvideo-avatar: High-fidelity audio-driven human animation for multiple characters

Audio-visual Controlled Video Diffusion with Masked Selective State Spaces Modelling for Natural Talking Head Generation

Hunyuanportrait: Implicit condition control for enhanced portrait animation

HunyuanVideo: A Systematic Framework For Large Video Generation Model

Multimodal Human Motion

AvatarGPT: All-in-One Framework for Motion Understanding, Planning, Generation and Beyond

A Unified Framework for Multimodal, Multi-Part Human Motion Synthesis

MDSC: Towards Evaluating the Style Consistency Between Music and Dance

UDE: A Unified Driving Engine for Human Motion Generation

Remote Sensing & Civil Engineering

Community-scale multi-level post-hurricane damage assessment of residential buildings using multi-temporal airborne LiDAR data

Automated residential building detection from airborne LiDAR data with deep neural networks

Automated analysis of mobile LiDAR data for component-level damage assessment of building structures during large coastal storm events

Building smart and accessible transportation hubs with internet of things, big data analytics, and affective computing

Building Smart Transportation Hubs with 3D Vision and Video Technologies to Improve Services to People with Disabilities

Multiresolution change analysis framework for postdisaster assessment of natural gas pipeline risk

Image-based 3D reconstruction for posthurricane residential building damage assessment