Holo-World: Unified Camera, Object and Weather Control for Video World Model

2026-06-18

AI Summary

Existing video world models struggle to jointly control camera motion, object dynamics, and weather conditions from a single image while relying on source videos that contain future structural information. This work proposes a first-frame-anchored source-to-state generation paradigm, introducing the unified control dataset HoloStateData and a parameter subspace disentanglement mechanism to preserve scene structure while enabling weather transfer. By integrating rendered backgrounds, geometric buffers, and object controls through a unified scene adapter, and employing a scene–weather disentangled classifier-free guidance strategy, the method achieves precise control over camera and object motion alongside diverse weather synthesis. Experiments demonstrate that the proposed approach outperforms existing video-to-video weather editing methods across both quantitative and qualitative metrics.

Abstract

Video world models are moving toward preserving an observed world under controllable camera and object motion while allowing its environmental state to change. Yet these controls remain isolated, and weather generation typically relies on a source video or reconstructed scene that already specifies future structure. We study a first-frame-anchored source-to-state setting, where the model starts from a single image and follows explicit camera and object controls and an optional weather instruction, then generates a video that either preserves the source world or transfers it to a target weather state. To address these challenges, we first build HoloStateData, a state video dataset that turns diverse videos into unified control samples for camera, object, and weather supervision. Second, we introduce Holo-World, a unified controllable video world model that jointly controls scene from a single image. Its Unified Scene Adapter factorizes world preservation and weather transfer into distinct parameter subspaces, using rendered background, geometry buffers, and object controls to maintain controlled scene structure while modeling weather-dependent appearance and particle effects. Additionally, Scene-Weather Decomposed CFG guides scene and weather residuals separately, strengthening target weather effects without over-amplifying the full condition. Quantitative and qualitative experiments demonstrate that Holo-World maintains precise camera and object control with consistent scene structure while transferring scenes into diverse target weather state, outperforming video-to-video weather editing baselines on weather-state generation. Our project page is available at \url{https://xiangchenyin.github.io/Holo-World/}.

Problem

• video world model
• camera control
• object control
• weather generation
• single-image-to-video

Innovation

• Unified Scene Adapter
• Scene-Weather Decomposed CFG
• Video World Model
• First-Frame-Anchored Generation
• Controllable Weather Transfer