| title | 4DAnyone: Create Anyone in 4D from a Casual Monocular Video |
|---|---|
| authors | Yudong Jin, Tao Xie, Qihang Zhang, Zehong Shen, Zhen Xu, Yujun Shen, Hujun Bao, Xiaowei Zhou, Yinghao Xu |
| date | 2026-08-20 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2608.20335 |
| affiliation | Robbyant Research |
| upvotes | 83 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | We present 4DAnyone, a framework for reconstructing 4D humans from an uncalibrated monocular video by generating reconstruction-grade multiview-consistent videos and lifting them into 4D Gaussian Splatting (4DGS). Existing camera-controlled video diffusion models synthesize plausible novel-view videos but fail to maintain consistency when scaled to the tens of target views required for 4DGS reconstruction. We identify this failure as a bounded-attention-context problem: when target views exceed the capacity of a single DiT forward pass, they must be split into groups, exposing two coupled bottlenecks. On the reference-context side, conditioning on all previously generated views grows as O(N), weakening cross-view appearance guidance. On the target-context side, disjoint groups cannot directly exchange information, causing global structural drift. 4DAnyone addresses both bottlenecks with two complementary designs: Reference Context Packing (RCP) compresses growing reference views into a fixed-length mixed-resolution context with O(1) reference-context complexity, while Target Context Routing (TCR) rotates target-view groupings during denoising to share context across groups at high-noise steps and stabilize details at low-noise steps. We further build the MVGameHuman dataset using our in-house game engine and combine it with light-stage and in-the-wild video datasets for training. Experiments on DNA-Rendering and DyMVHumans show that 4DAnyone outperforms prior methods in both novel-view video quality and downstream 4DGS reconstruction, with robust in-the-wild generalization. See our project page for video results and source code: https://4danyone.github.io. |
cite
@misc{jin2026danyone,
title={4DAnyone: Create Anyone in 4D from a Casual Monocular Video},
author={Yudong Jin and Tao Xie and Qihang Zhang and Zehong Shen and Zhen Xu and Yujun Shen and Hujun Bao and Xiaowei Zhou and Yinghao Xu},
year={2026},
eprint={2608.20335},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2608.20335}
}