mlsanityData to 14 Oct 2026

ESG: Generating Physically Consistent Dynamic 3D Scenes from Text Descriptions

arXiv 2609.15392 ↗First posted 14 Sep 2026cs.GR cs.CVcited by 0

Not on the map. Why?

Abstract (arXiv)

Recent progress in image and 3D scene generation has enabled increasingly realistic static environments, yet most methods remain confined to such static configurations. Generating dynamic scenes from natural language is fundamentally challenging: it requires joint reasoning over scene structure, temporal evolution, and physical feasibility, while ensuring reliable execution in modern physics engines. We present a unified framework for generating physically consistent dynamic 3D scenes from text, with outputs directly executable in Unreal Engine. Central to our approach is the Evolutive Scene Graph (ESG), which specifies entities with physical attributes, spatial relations, and event-driven timelines in a machine-checkable form. Given a prompt, a large language model constructs and validates a complete ESG; spatial layouts are grounded via energy-minimized gradient optimization; timeline-constrained physical parameters are then optimized through differentiable simulation to satisfy user-specified events; and the resulting scene is compiled into an engine-executable class. Experiments on 10 scenes across three complexity levels show that our method achieves 16.4/18 mean event completion, outperforming Scene Language, the strongest engine-executable baseline (SimWorld), and our ablation without physical optimization by a clear margin in event completion and parameter accuracy.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 39 references that are in the map. First posted on arXiv 14 Sep 2026.

Builds on

27 of 39 references.

References · 27

  1. 2024Structured 3D Latents for Scalable and Versatile 3D Generationcited by 770
  2. 20203D-FRONT: 3D Furnished Rooms with layOuts and semaNTicscited by 380
  3. 2023Holodeck: Language Guided Generation of 3D Embodied AI Environmentscited by 268
  4. 2019DiffTaichi: Differentiable Programming for Physical Simulationcited by 258
  5. 2021ATISS: Autoregressive Transformers for Indoor Scene Synthesiscited by 200
  6. 2023DiffuScene: Denoising Diffusion Models for Generative Indoor Scene Synthesiscited by 166
  7. 2020SceneFormer: Indoor Scene Generation with Transformerscited by 158
  8. 2023PAC-NeRF: Physics Augmented Continuum Neural Radiance Fields for Geometry-Agnostic System Identificationcited by 154
  9. 2024SceneCraft: An LLM Agent for Synthesizing 3D Scene as Blender Codecited by 134
  10. 2021Unconstrained Scene Generation with Locally Conditioned Radiance Fieldscited by 130
  11. 2024InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Priorcited by 117
  12. 2018Fast and Flexible Indoor Scene Synthesis via Deep Convolutional Generative Modelscited by 106
15 more
  1. 2021Graph-to-3D: End-to-End Generation and Manipulation of 3D Scenes Using Scene Graphscited by 84
  2. 2026WorldArena: A Unified Benchmark for Evaluating Perception and Functional Utility of Embodied World Modelscited by 76
  3. 2021DiffPD: Differentiable Projective Dynamicscited by 62
  4. 2023GPT4Motion: Scripting Physical Motions in Text-to-Video Generation via Blender-Oriented GPT Planningcited by 61
  5. 2025SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agentcited by 56
  6. 2019SceneGraphNet: Neural Message Passing for 3D Indoor Scene Augmentationcited by 55
  7. 2023CommonScenes: Generating Commonsense 3D Indoor Scenes with Scene Graph Diffusioncited by 54
  8. 2022NeuPhysics: Editable Neural Geometry and Physics from Monocular Videoscited by 50
  9. 2024Director3D: Real-world Camera Trajectory and 3D Scene Generation from Textcited by 50
  10. 2024SceneCraft: Layout-Guided 3D Scene Generationcited by 41
  11. 2024The Scene Language: Representing Scenes with Programs, Words, and Embeddingscited by 36
  12. 2025DSO: Aligning 3D Generators with Simulation Feedback for Physical Soundnesscited by 25
  13. 2025PAT3D: Physics-Augmented Text-to-3D Scene Generationcited by 16
  14. 2025SimWorld: An Open-ended Realistic Simulator for Autonomous Agents in Physical and Social Worldscited by 12
  15. 2026FATE: Closed-Loop Feasibility-Aware Task Generation with Active Repair for Physically Grounded Robotic Curriculacited by 3

Built on by

Not cited yet.