Voyager: Long-Range and World-Consistent Video Diffusion for Explorable 3D Scene Generation

arXiv 25.6 2025 46.6 method

TLDR

Voyager generates explorable 3D scenes from a single image using world-consistent video diffusion with a cache for long-range coherence.

Reasoning

The paper presents a novel video diffusion framework for 3D scene generation, with strengths in end-to-end consistency and long-range exploration. Weaknesses include lack of explicit real-world evaluation and limited connection to reinforcement learning or dynamics prediction.

Read-first score

Read-first score 46.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 25.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=metric,result

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=github

Topical relevance 42%

35.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 372.

Keyword Scores

video world model

interactive world model

world model

generative world model

world simulator

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

World-Consistent Video Diffusion: A unified architecture that jointly generates aligned RGB and depth video sequences, conditioned on existing world observation to ensure global coherence.
Long-Range World Exploration: An efficient world cache with point culling and an auto-regressive inference with smooth video sampling for iterative scene extension with context-aware consistency.
Scalable Data Engine: A video reconstruction pipeline that automates camera pose estimation and metric depth prediction for arbitrary videos, enabling large-scale, diverse training data curation without manual 3D annotations.

Methodology

Voyager is a video diffusion framework that generates world-consistent 3D point-cloud sequences from a single image with user-defined camera path. It integrates three key components: world-consistent video diffusion for joint RGB and depth generation, long-range world exploration using a world cache and auto-regressive inference, and a scalable data engine for automated pose and depth estimation. The method achieves end-to-end scene generation and reconstruction without requiring traditional 3D reconstruction pipelines like structure-from-motion or multi-view stereo.

Key Results

Voyager demonstrates clear improvement over existing methods in visual quality and geometric accuracy, with versatile applications for explorable 3D scene generation.

Resources

Paper Project