mlsanityData to 7 Oct 2026

Persistent Nature: A Generative Model of Unbounded 3D Worlds

arXiv 2303.13515 ↗First posted 23 Mar 2023cs.CV cs.LGcited by 44

On the map: World models → Video-generation world models and simulators

Abstract (arXiv)

Despite increasingly realistic image quality, recent 3D image generative models often operate on 3D volumes of fixed extent with limited camera motions. We investigate the task of unconditionally synthesizing unbounded nature scenes, enabling arbitrarily large camera motion while maintaining a persistent 3D world model. Our scene representation consists of an extendable, planar scene layout grid, which can be rendered from arbitrary camera poses via a 3D decoder and volume rendering, and a panoramic skydome. Based on this representation, we learn a generative world model solely from single-view internet photos. Our method enables simulating long flights through 3D landscapes, while maintaining global scene consistency--for instance, returning to the starting point yields the same view of the scene. Our approach enables scene extrapolation beyond the fixed bounds of current 3D generative models, while also supporting a persistent, camera-independent world representation that stands in contrast to auto-regressive 3D prediction models. Our project page: https://chail.github.io/persistent-nature/.

Key facts. Cited by 44 papers in the map (data through 7 Oct 2026). Placed in Video-generation world models and simulators, one of 11 directions in World models. Builds on 85 of its 105 references that are in the map. First posted on arXiv 23 Mar 2023.

Builds on

85 of 105 references.

Closest ancestors · 2

Same direction, newest first

  1. 2022DreamFusion: Text-to-3D using 2D Diffusion29 Sep
  2. 2021Learning Transferable Visual Models From Natural Language Supervision26 Feb

Other references · 83

  1. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  2. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,580
  3. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,013
  4. 2021Vision Transformers for Dense Predictioncited by 1,789
  5. 2022Imagen Video: High Definition Video Generation with Diffusion Modelscited by 1,765
  6. 2022Make-A-Video: Text-to-Video Generation without Text-Video Datacited by 1,725
  7. 2021Plenoxels: Radiance Fields without Neural Networkscited by 1,607
  8. 2020pixelNeRF: Neural Radiance Fields from One or Few Imagescited by 1,574
  9. 2021Palette: Image-to-Image Diffusion Modelscited by 1,296
  10. 2021Efficient Geometry-aware 3D Generative Adversarial Networkscited by 1,204
  11. 2021Alias-Free Generative Adversarial Networkscited by 1,181
  12. 2020GIRAFFE: Representing Scenes as Compositional Generative Neural Feature Fieldscited by 784
38 more
  1. 2018Recurrent World Models Facilitate Policy Evolutioncited by 761
  2. 2021Resolution-robust Large Mask Inpainting with Fourier Convolutionscited by 730
  3. 2020pi-GAN: Periodic Implicit Generative Adversarial Networks for 3D-Aware Image Synthesiscited by 708
  4. 2021IBRNet: Learning Multi-View Image-Based Renderingcited by 698
  5. 2020GRAF: Generative Radiance Fields for 3D-Aware Image Synthesiscited by 688
  6. 2018Which Training Methods for GANs do actually Converge?cited by 672
  7. 2019Local Light Field Fusion: Practical View Synthesis with Prescriptive Sampling Guidelinescited by 672
  8. 2020GANSpace: Discovering Interpretable GAN Controlscited by 612
  9. 2018Free-Form Image Inpainting with Gated Convolutioncited by 594
  10. 2018Generative Image Inpainting with Contextual Attentioncited by 578
  11. 2017MoCoGAN: Decomposing Motion and Content for Video Generationcited by 523
  12. 2021StyleNeRF: A Style-based 3D-Aware Generator for High-resolution Image Synthesiscited by 477
  13. 2020Closed-Form Factorization of Latent Semantics in GANscited by 407
  14. 2016Generating Videos with Scene Dynamicscited by 376
  15. 2019SynSin: End-to-end View Synthesis from a Single Imagecited by 364
  16. 2021StyleGAN-V: A Continuous Video Generator with the Price, Image Quality and Perks of StyleGAN2cited by 300
  17. 2021StyleSDF: High-Resolution 3D-Consistent Image and Geometry Generationcited by 289
  18. 2020Single-View View Synthesis with Multiplane Imagescited by 263
  19. 2016Unsupervised Learning for Physical Interaction through Video Predictioncited by 256
  20. 2019DeepView: View Synthesis with Learned Gradient Descentcited by 253
  21. 2019On the "steerability" of generative adversarial networkscited by 247
  22. 2018Stochastic Video Generation with a Learned Priorcited by 244
  23. 2022Long Video Generation with Time-Agnostic VQGAN and Time-Sensitive Transformercited by 244
  24. 2020Infinite Nature: Perpetual View Generation of Natural Scenes from a Single Imagecited by 239
  25. 2020Free View Synthesiscited by 231
  26. 2021Large Scale Image Completion via Co-Modulated Generative Adversarial Networkscited by 219
  27. 2021GRAM: Generative Radiance Manifolds for 3D-Aware Image Generationcited by 216
  28. 2022MAT: Mask-Aware Transformer for Large Hole Image Inpaintingcited by 216
  29. 2018Stereo Magnification: Learning View Synthesis using Multiplane Imagescited by 206
  30. 20203D Photography using Context-aware Layered Depth Inpaintingcited by 206
  31. 2022Generating Videos with Dynamics-aware Implicit Generative Adversarial Networkscited by 186
  32. 2021CodeNeRF: Disentangled Neural Radiance Fields for Object Categoriescited by 183
  33. 2019Neural Volumes: Learning Dynamic Renderable Volumes from Imagescited by 173
  34. 2017Decomposing Motion and Content for Natural Video Sequence Predictioncited by 168
  35. 2023SceneScape: Text-Driven Consistent Scene Generationcited by 160
  36. 2021MINE: Towards Continuous Depth MPI with NeRF for Novel View Synthesiscited by 133
  37. 2021Unconstrained Scene Generation with Locally Conditioned Radiance Fieldscited by 130
  38. 20193D Ken Burns Effect from a Single Imagecited by 128

50 most cited of 83 All 83 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 6; 2024: 13; 2025: 21; 2026: 4.

44 papers.

Sort:

By area:

  1. 2024WonderWorld: Interactive 3D Scene Generation from a Single Imagecited by 197
  2. 2023WonderJourney: Going from Anywhere to Everywherecited by 157
  3. 2025CameraCtrl II: Dynamic Scene Exploration via Camera-controlled Video Diffusion Modelscited by 97
  4. 2023CityDreamer: Compositional Generative Model of Unbounded 3D Citiescited by 75
  5. 2025WoW: Towards a World omniscient World model Through Embodied Interactioncited by 66
  6. 2023NeRFiller: Completing Scenes via Generative 3D Inpaintingcited by 62
  7. 2024Generative Gaussian Splatting for Unbounded 3D City Generationcited by 43
  8. 2024Streetscapes: Large-scale Consistent Street View Generation Using Autoregressive Video Diffusioncited by 39
  9. 2025SynCity: Training-Free Generation of 3D Worldscited by 30
  10. 2025Learning 3D Persistent Embodied World Modelscited by 29
  11. 20253D Scene Generation: A Surveycited by 24
  12. 2024UrbanWorld: An Urban World Model for 3D City Generationcited by 23
32 more
  1. 2023CityGen: Infinite and Controllable City Layout Generationcited by 22
  2. 2023iNVS: Repurposing Diffusion Inpainters for Novel View Synthesiscited by 20
  3. 2024CityX: Controllable Procedural Content Generation for Unbounded 3D Citiescited by 20
  4. 2023Make-It-4D: Synthesizing a Consistent Long-Term Dynamic Scene Video from a Single Imagecited by 19
  5. 2025NuiScene: Exploring Efficient Generation of Unbounded Outdoor Scenescited by 19
  6. 2025Compositional Generative Model of Unbounded 4D Citiescited by 16
  7. 2025Sat2City: 3D City Generation from A Single Satellite Image with Cascaded Latent Diffusioncited by 15
  8. 2024Urban Scene Diffusion through Semantic Occupancy Mapcited by 14
  9. 2024COHO: Context-Sensitive City-Scale Hierarchical Urban Layout Generationcited by 10
  10. 2025Learning-based 3D Reconstruction in Autonomous Driving: A Comprehensive Surveycited by 9
  11. 2025Yo'City: Personalized and Boundless 3D Realistic City Scene Generation via Self-Critic Expansioncited by 8
  12. 2025ScenePainter: Semantically Consistent Perpetual 3D Scene Generation with Concept Relation Alignmentcited by 7
  13. 2025TRELLISWorld: Training-Free World Generation from Object Generatorscited by 7
  14. 2024Proc-GS: Procedural Building Generation for City Assembly with 3D Gaussianscited by 6
  15. 2026WorldClaw: Agentic 3D Open-World Generation at Scalecited by 6
  16. 2024Locally Adaptive Neural 3D Morphable Modelscited by 5
  17. 20246Img-to-3D: Few-Image Large-Scale Outdoor Driving Scene Reconstructioncited by 5
  18. 2025MajutsuCity: Language-driven Aesthetic-adaptive City Generation with Controllable 3D Assets and Layoutscited by 5
  19. 2025MeSS: City Mesh-Guided Outdoor Scene Generation with Cross-View Consistent Diffusioncited by 3
  20. 2025Look Beyond: Two-Stage Scene View Generation via Panorama and Video Diffusioncited by 3
  21. 2025Self-Evolving 3D Scene Generation from a Single Imagecited by 3
  22. 2024MaGRITTe: Manipulative and Generative 3D Realization from Image, Topview and Textcited by 2
  23. 2024iControl3D: An Interactive System for Controllable 3D Scene Generationcited by 2
  24. 2024EarthGen: Generating the World from Top-Down Viewscited by 2
  25. 2025Constructing a 3D Scene from a Single Imagecited by 2
  26. 2026Sat2City v2: Native 3D City Asset Generation from a Single Satellite Imagecited by 2
  27. 2025DreamJourney: Perpetual View Generation with Video Diffusion Modelscited by 1
  28. 2025NeoWorld: Neural Simulation of Explorable Virtual Worlds via Progressive 3D Unfoldingcited by 1
  29. 2025WonderZoom: Multi-Scale 3D World Generationcited by 1
  30. 2026DepthART: Scaling Foundation Monocular Depth to Tiny Modelscited by 1
  31. 2025WonderVerse: Extendable 3D Scene Generation with Video Generative Modelscited by 0
  32. 2026WorldWeave: Growing Persistent Geometric Worlds for Video Generationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Beyond Pixel Histories: World Models with Persistent 3D State
  2. 2025WorldGrow: Generating Infinite 3D World
  3. 2026WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation
  4. 2025WorldReel: 4D Video Generation with Consistent Geometry and Motion Modeling
  5. 2026Future Dynamic 3D Reconstruction: Toward 3D World Modeling with Disentangled Ego-Motion
  6. 2025GEN3D: Generating Domain-Free 3D Scenes from a Single Image
  7. 2026LiveWorld: Simulating Out-of-Sight Dynamics in Generative Video World Models
  8. 2025Terra: Explorable Native 3D World Model with Point Latents