| title | Geo-Align: Video Generation Alignment via Metric Geometry Reward |
|---|---|
| authors | Zizun Li, Haoyu Guo, Runzhe Teng, Chunhua Shen, Tong He |
| date | 2026-05-22 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2605.23903 |
| upvotes | 7 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Camera-controlled video generation has achieved remarkable progress in recent years. However, existing video-to-video re-rendering methods primarily rely on Supervised Fine-Tuning using synthetic datasets. At present, there is an extreme scarcity of synchronized, multi-view real-world video data. Consequently, the prevailing paradigm often exhibits limited generalization when processing out-of-distribution real-world videos, with models struggling to accurately adhere to physical scales and camera trajectories. To bridge this gap, we propose Geo-Align, the first Reinforcement Learning framework specifically designed for camera-controlled video re-rendering. Built upon a pretrained model, we optimize the model through a scale-aware perceptual reward mechanism. Specifically, we introduce a metric 3D estimator to extract precise camera trajectories from generated videos, explicitly penalizing deviations in rotation and translation. Furthermore, we meticulously designed a data pipeline strategy based on real-world conditioning videos and target camera trajectories derived from synthetic data, eliminating the reliance on paired data. Extensive experiments demonstrate that Geo-Align consistently outperforms existing supervised learning baselines in both precise camera controllability and visual fidelity, indicating the effectiveness of our method. |
cite
@misc{li2026geo,
title={Geo-Align: Video Generation Alignment via Metric Geometry Reward},
author={Zizun Li and Haoyu Guo and Runzhe Teng and Chunhua Shen and Tong He},
year={2026},
eprint={2605.23903},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2605.23903}
}