|
I'm Boyang Wang (ēåę“), and I work on video generation. I believe visual context compression, especially for video, is a key source of intelligence for video models. I'm currently a second-year CS PhD student at the University of Virginia (UVA), supervised by Zezhou Cheng. Before joining UVA, I earned an M.S. in Robotics and a B.S. in Computer Science from the University of Michigan. At UMich, I worked with Prof. JJ Park, Prof. Nima Fazeli, Prof. Hun-Seok Kim, and Prof. Chad Jenkins. Since the summer of 2026, I have been interning at Adobe, where I am fortunate to be mentored by Matheus Gadelha. During the summer of 2025, I interned at Shanghai AI Lab, working closely with Xudong Xu in the Embodied AI Department. My research interests mainly lie in Generative Models (especially video) for General, Robotics, and Anime domains! I am actively seeking research internship opportunities for Summer 2027 in generative models, especially video generation. Feel free to reach out! Email Ā· Twitter Ā· Google Scholar Ā· GitHub Ā· Hugging Face |
Taken in Northern Michigan, 2024 |
|
|
|
Show older news |
|
|
|
Shengxiang Ji, Boyang Wang, Haiyang Xu, Bingnan Li, Yucheng Mao, Zeyuan Chen, Xiaojun Shan, Xiang Zhang, Gang Hua, Jianwen Xie, Zezhou Cheng, Zhuowen Tu arXiv 2026 Camera-controlled image-to-video generation under large viewpoint changes, where a last-frame layout specifies what should appear in the future view and where. |
|
|
Boyang Wang, Guangyi Xu, Zhipeng Tang, Jiahui Zhang, Zezhou Cheng NeurIPS 2026 State-of-the-art shot boundary detection that is more sensitive and more informative.
25k+ model downloads on Hugging Face Ā· 340+ GitHub stars
|
|
|
Boyang Wang*, Haoran Zhang*, Shujie Zhang*, Jinkun Hao, Mingda Jia, Qi Lv, Yucheng Mao, Zhaoyang Lyu, Jia Zeng, Xudong Xu, Jiangmiao Pang arXiv 2026 A multi-view, inpainting-based video diffusion model conditioned on identity references, which augments robot manipulation data in both simulation and real-robot setups. |
|
Boyang Wang, Xuweiyi Chen, Matheus Gadelha, Zezhou Cheng NeurIPS 2025 The first video generation work to study the Frame In and Frame Out cinematic effects, conditioned on an expanded first-frame canvas, a text prompt, motion trajectories, and an identity reference. 2.4k+ model downloads on Hugging Face |
|
Yucheng Mao*, Boyang Wang*, Nilesh Kulkarni, Jeong Joon Park CVPR 2025 Reconstructs a 3D scene from degraded image sets with a multi-view diffusion model. |
|
Boyang Wang, Nikhil Sridhar, Chao Feng, Mark Van der Merwe, Adam Fishman, Nima Fazeli, Jeong Joon Park ICRA 2025 The first video generative model conditioned on both language and gestures, used to guide robot actions. |
|
Boyang Wang, Fengyu Yang, Xihang Yu, Chao Zhang, Hanbin Zhao CVPR 2024 State-of-the-art anime super-resolution and restoration.
22k+ model weight downloads Ā· 1.1k+ GitHub stars Ā· 140+ likes on the Hugging Face demo
|
|
Boyang Wang*, Bowen Liu*, Shiyu Liu*, Fengyu Yang WACV 2024 The first work to study video compression degradation for image super-resolution. |
|
|
|
Research Intern, 05/2026 - Present Mentor: Matheus Gadelha Research on efficient all-in-one video generation. |
|
Research Intern, Embodied AI Department, 05/2025 - 08/2025 Mentor: Xudong Xu Multi-view video generation to augment robot manipulation data, which led to RoboVIP. |
|
Machine Learning Engineer, 04/2022 - 08/2022 Engineering work on video compression codecs. |
|
|
|
|
|
|
|
|
|
|
|
I believe that the optimizer of life is the faith one holds. |