Geo-Align: Video Generation Alignment via Metric Geometry Reward

Hugging Face Daily Papers Papers

Summary

Geo-Align presents a reinforcement learning framework for camera-controlled video re-rendering that improves generalization through scale-aware perceptual rewards and metric 3D estimation for camera trajectory extraction.

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.
Original Article
View Cached Full Text

Cached at: 05/25/26, 06:36 AM

Paper page - Geo-Align: Video Generation Alignment via Metric Geometry Reward

Source: https://huggingface.co/papers/2605.23903

Abstract

Geo-Align presents a reinforcement learning framework for camera-controlled video re-rendering that improves generalization through scale-aware perceptual rewards and metric 3D estimation for camera trajectory extraction.

Camera-controlled video generation has achieved remarkable progress in recent years. However, existing video-to-video re-rendering methods primarily rely onSupervised Fine-Tuningusingsynthetic datasets. At present, there is an extreme scarcity of synchronized, multi-viewreal-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 andcamera trajectories. To bridge this gap, we propose Geo-Align, the firstReinforcement Learningframework specifically designed forcamera-controlled video re-rendering. Built upon apretrained model, we optimize the model through ascale-aware perceptual rewardmechanism. Specifically, we introduce ametric 3D estimatorto extract precisecamera trajectoriesfrom generated videos, explicitly penalizing deviations in rotation and translation. Furthermore, we meticulously designed adata pipeline strategybased on real-world conditioning videos and targetcamera trajectoriesderived 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.

View arXiv pageView PDFProject pageGitHub3Add to collection

Get this paper in your agent:

hf papers read 2605\.23903

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.23903 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.23903 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.23903 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Towards Consistent Video Geometry Estimation

Hugging Face Daily Papers

ViGeo is a transformer-based foundation model that recovers dense and consistent 3D geometry from videos using dynamic chunking attention and a completion-based data refinement framework, achieving state-of-the-art performance across multiple tasks.