Skip to content

3D Scene / Environment Generation

3D Generation

Open Source

  • DW

    DreamX-World 1.0 Discovered by AI

    2026

    A general-purpose interactive text/image-to-video world model for controllable long-horizon 3D scene generation. It supports camera navigation, revisits to previously observed regions, and promptable events across photorealistic, game-style, and stylized domains.

    interactive scene-generation world-model

    GitHub · arXiv

  • G3

    Gen3R: 3D Scene Generation Meets Feed-Forward Reconstruction Discovered by AI

    2026

    A method that bridges foundational reconstruction models and video diffusion models to generate both RGB videos and corresponding 3D geometry (including camera trajectories) for scene-level 3D generation.

    3d-scene reconstruction video-diffusion

    GitHub · arXiv

  • L2

    Lyra 2.0 Discovered by AI

    2026

    Developed by NVIDIA, Lyra 2.0 is a framework for generating persistent, explorable 3D worlds at scale from a single image or video walkthrough. It addresses spatial forgetting and temporal drifting over long camera trajectories to produce 3D-consistent environments ready for real-time rendering.

    3d-scene consistency video-to-3d world-model

    GitHub · arXiv · Paper · Website

  • MI

    minWM Discovered by AI

    2026

    A full-stack open-source framework and tutorial designed for real-time interactive video world models. It provides a complete pipeline spanning data construction, controllable fine-tuning, autoregressive training, few-step distillation, and streaming inference.

    interactive real-time world-model

    GitHub · arXiv

  • OA

    OpenWorldLib: A Unified Codebase and Definition of Advanced World Models Discovered by AI

    2026

    A unified world model inference framework that standardizes tasks such as interactive video generation, 3D scene generation, multimodal reasoning, and vision-language-action (VLA) models.

    3d-scene framework world-model

    GitHub · arXiv

  • PU

    PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Space Discovered by AI

    2026

    A unified framework for 3D scene generation and reconstruction operating directly in pixel space. By avoiding the information loss of latent space encoding, it achieves high-fidelity 3D scene synthesis and reconstruction without requiring a pretrained VAE.

    environment pixel-space reconstruction scene

    GitHub · arXiv

  • SA

    SAGE Discovered by AI

    2026

    An agentic framework by NVIDIA that automatically generates simulation-ready, realistic, and physically stable 3D environments from arbitrary user prompts. It couples layout and object generators with critics evaluating semantic plausibility, visual realism, and physical stability.

    agentic embodied-ai layout scene-generation

    Project · GitHub · arXiv

  • SA

    SceneAssistant: A Visual Feedback Agent for Open-Vocabulary 3D Scene Generation Discovered by AI

    2026

    A vision-feedback-driven agentic framework designed for open-vocabulary 3D scene synthesis, utilizing a suite of Action APIs to iteratively refine scene layouts.

    3d-generation agentic open-vocabulary scene-generation

    GitHub · arXiv

  • SG

    SceneFoundry: Generating Interactive Infinite 3D Worlds Discovered by AI

    2026

    A language-guided diffusion framework that generates apartment-scale 3D environments with functionally articulated furniture and semantically diverse layouts. It uses an LLM for floor plan generation and diffusion-based placement to ensure physical usability and navigability.

    embodied-ai environment interactive scene

    GitHub · arXiv

  • S3

    SynCity 3000 Discovered by AI

    2026

    A framework for generating globally coherent 3D scenes with fine-grained layout control by adapting image-to-3D generators as convolutional operators. It addresses the lack of scene-scale training data by fine-tuning on a novel synthetic data engine.

    3d-diffusion scene-scale synthetic-data

    GitHub · arXiv

  • WA

    WorldClaw: Agentic 3D Open-World Generation at Scale Discovered by AI

    2026

    An agentic coarse-to-fine framework for open-world 3D scene generation. It uses planning agents to translate text prompts into structured specifications of regions, terrain, assets, materials, and spatial relations, and then builds a globally coherent 3D scene.

    agentic environment open-world scene

    Project · GitHub · arXiv

  • WF

    WorldFlow3D: Flowing Through 3D Distributions for Unbounded World Generation Discovered by AI

    2026

    A framework for generating unbounded 3D worlds with full controllability using flow matching to model 3D data distributions, enabling large-scale outdoor and indoor scene synthesis.

    3d-generation flow-matching scene-generation unbounded-world

    GitHub · arXiv

  • BL

    BloomScene Discovered by AI

    2025

    A lightweight structured 3D Gaussian Splatting framework for crossmodal scene generation from text or image inputs. It addresses storage overhead and geometric distortion issues in 3D scene generation by introducing a structured Gaussian representation.

    3dgs image-to-3d scene-generation text-to-3d

    GitHub · arXiv

  • BO

    Bolt3D Discovered by AI

    2025

    A latent diffusion model designed for fast, feed-forward 3D scene generation that directly samples a 3D scene representation in less than seven seconds on a single GPU. It leverages scalable 2D diffusion network architectures to produce consistent, high-fidelity multi-view Splatter Images that combine into a complete 3D Gaussian scene.

    diffusion fast-generation feed-forward gaussian-splatting

    arXiv · Paper · Website

  • FL

    FlashWorld Discovered by AI

    2025

    A fast and high-quality 3D scene generation model that produces 3D scenes from a single image or text prompt in seconds. It uses a cross-mode post-training distillation strategy to transfer high visual fidelity from a multi-view-oriented diffusion model to a 3D-consistent multi-view generative model.

    3dgs fast-generation image-to-3d text-to-3d

    Project · GitHub · arXiv

  • FM

    FreeScene: Mixed Graph Diffusion for 3D Scene Synthesis from Free Prompts Discovered by AI

    2025

    A user-friendly controllable indoor scene synthesis framework that integrates a vision-language model-based graph designer with a generative Mixed Graph Diffusion Transformer to support free-form user inputs like text, diagrams, and sketches.

    3d-scene diffusion indoor-synthesis

    GitHub · arXiv

  • H2

    HOLODECK 2.0: Vision-Language-Guided 3D World Generation with Editing Discovered by AI

    2025

    An advanced vision-language-guided framework for 3D world generation that supports interactive scene editing based on human feedback, enabling the creation of stylistically varied 3D environments.

    3d-scene diffusion environment

    GitHub · arXiv

  • H2

    HunyuanWorld 2.0 (HY-World-2.0)

    2025

    Tencent Hunyuan's open 3D world generation model that creates explorable, immersive 3D scenes and worlds from text or image prompts.

    open-weights scene-generation world-model

    GitHub

  • LG

    Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation Discovered by AI

    2025

    A generative framework for 3D and 4D scene reconstruction that distills implicit 3D knowledge from camera-controlled video diffusion models into explicit 3D Gaussian Splatting representations. It supports generating explorable 3D environments from text, single images, or monocular videos.

    3d-scene gaussian-splatting video-diffusion world-model

    Project · GitHub · arXiv

  • MM

    MAGIC: Multi-scene Automated Gameworlds Generation with Intelligent Connectivity Discovered by AI

    2025

    A framework for multi-scene automated gameworld generation with intelligent connectivity using large language models, supporting scene planning, composition, and integration with Unity and Blender.

    blender gameworld scene unity

    GitHub

  • M3

    Matrix-3D: Omnidirectional Explorable 3D World Generation Discovered by AI

    2025

    A framework that utilizes panoramic representation for wide-coverage omnidirectional explorable 3D world generation, combining conditional video generation and panoramic reconstruction.

    3d-scene panorama world-generation

    GitHub · arXiv

  • MM

    MMGDreamer: Mixed-Modality Graph for Geometry-Controllable 3D Indoor Scene Generation Discovered by AI

    2025

    A dual-branch diffusion framework for controllable 3D indoor scene generation that utilizes mixed-modality graphs, visual enhancement, and relation prediction to achieve precise control over object geometry.

    3d-scene diffusion indoor-synthesis

    GitHub · arXiv

  • SS

    Scene Splatter: Momentum 3D Scene Generation from Single Image with Video Diffusion Model Discovered by AI

    2025

    A momentum-based paradigm for video diffusion that generates generic 3D scenes from a single image. It addresses the limitations of video length and scene inconsistency in novel-view synthesis by constructing noisy samples from historical frames, enabling stable and high-quality 3D reconstruction.

    diffusion gaussian-splatting scene-generation video-diffusion

    Project · GitHub · arXiv

  • SC

    SceneGen Discovered by AI

    2025

    A feedforward framework for generating multiple 3D assets from a single scene image and its corresponding object masks. It simultaneously produces geometry, texture, and relative spatial positions for multiple objects without requiring optimization or asset retrieval.

    feedforward image-to-3d multi-object scene-generation

    Project · GitHub · arXiv

  • SS

    ScenePainter: Semantically Consistent Perpetual 3D Scene Generation with Concept Relation Alignment Discovered by AI

    2025

    A framework for perpetual 3D scene generation that produces long-range, coherent 3D view sequences from a single image. It addresses semantic drift by aligning the outpainting model's scene-specific priors with concept relations represented in a scene concept graph.

    outpainting perpetual-scene scene-generation scene-graph

    Project · GitHub · arXiv

  • SA

    Scenethesis: A Language and Vision Agentic Framework for 3D Scene Generation Discovered by AI

    2025

    A training-free agentic framework that couples LLM-based scene planning with vision-guided layout refinement to generate diverse, realistic, and physically plausible 3D interactive scenes. It uses an optimization module to enforce accurate pose alignment and physical stability.

    agent environment layout scene

    Project · arXiv

  • SP

    SpatialGen Discovered by AI

    2025

    A layout-guided 3D indoor scene generation framework that leverages a multi-view, multi-modal diffusion model to synthesize photo-realistic 3D scenes from text or image prompts.

    diffusion indoor-scene layout-guided

    Project · GitHub · arXiv

  • WO

    WorldGrow

    2025

    Generates large, coherent 3D scenes by progressively growing and extending generated geometry beyond the initially observed region.

    3d-scene diffusion scene-generation

    Project · GitHub · arXiv

  • 3S

    3D-SceneDreamer Discovered by AI

    2024

    A text-driven 3D-consistent scene generation framework that addresses error accumulation in geometry and appearance by leveraging generative models to create consistent 3D environments.

    consistency scene-generation text-to-3d

    GitHub · arXiv

  • BE

    BlockFusion: Expandable 3D Scene Generation using Latent Tri-plane Extrapolation Discovered by AI

    2024

    A diffusion-based model that generates 3D scenes as unit blocks and seamlessly incorporates new blocks to extend the scene infinitely. It utilizes a latent tri-plane representation and an extrapolation mechanism to ensure geometric and stylistic consistency across block boundaries.

    diffusion infinite-world scene-generation tri-plane

    Project · GitHub · arXiv

  • DI

    Director3D Discovered by AI

    2024

    A text-to-3D generation framework designed for real-world datasets, generating both realistic 3D scenes using 3D Gaussian Splatting and camera trajectories from text prompts.

    gaussian-splatting scene-generation text-to-3d

    GitHub · arXiv

  • DR

    DreamScene360 Discovered by AI

    2024

    A text-to-3D scene generation pipeline that creates immersive 360-degree environments by lifting flat panoramic images into 3D Gaussian Splatting representations with semantic and geometric constraints.

    3d-scene gaussian-splatting panorama

    GitHub · arXiv

  • D3

    DreamScene: 3D Gaussian-based Text-to-3D Scene Generation via Formation Pattern Sampling Discovered by AI

    2024

    A 3D Gaussian Splatting-based text-to-3D scene generation framework that utilizes Formation Pattern Sampling (FPS) and a progressive camera sampling strategy to produce high-quality, consistent, and editable indoor and outdoor environments.

    3d-scene gaussian-splatting text-to-3d

    Project · GitHub · arXiv

  • EI

    EchoScene: Indoor Scene Generation via Information Echo over Scene Graph Diffusion Discovered by AI

    2024

    A generative model for interactive and controllable 3D indoor scene generation based on scene graphs. It uses a dual-branch diffusion model with an information echo scheme to coordinate shape and layout generation under global constraints.

    3d-scene diffusion-model scene-graph

    GitHub · arXiv · Website

  • FG

    Frankenstein: Generating Semantic-Compositional 3D Scenes in One Tri-Plane Discovered by AI

    2024

    A diffusion-based framework that generates semantic-compositional 3D scenes in a single pass. It decodes multiple Signed Distance Fields (SDFs) from a single tri-plane tensor, allowing the simultaneous generation of multiple separated, semantically meaningful shapes that can be individually edited or rearranged.

    compositional diffusion scene-generation sdf

    Project · GitHub · arXiv

  • II

    InstructScene: Instruction-Driven 3D Indoor Scene Synthesis with Semantic Graph Prior Discovered by AI

    2024

    A generative framework that synthesizes 3D indoor scenes from natural language instructions by integrating a semantic graph prior and a layout decoder to improve controllability and fidelity.

    3d-scene diffusion indoor-synthesis

    GitHub · arXiv

  • MI

    MIDI Discovered by AI

    2024

    A multi-instance diffusion model designed for single-image to compositional 3D scene generation, enabling the simultaneous generation of multiple high-quality 3D objects with accurate spatial relationships.

    3d-scene diffusion image-to-3d multi-instance

    Project · GitHub · arXiv

  • PP

    PhyScene: Physically Interactable 3D Scene Synthesis for Embodied AI Discovered by AI

    2024

    A guided diffusion model for physically interactable 3D scene synthesis, incorporating collision, layout, and reachability constraints to generate realistic layouts with articulated objects.

    3d-generation diffusion-model physics-constraints scene-generation

    GitHub · arXiv

  • RE

    RealmDreamer Discovered by AI

    2024

    A technique for generating general forward-facing 3D scenes from text descriptions by optimizing a 3D Gaussian Splatting representation using pretrained 2D image-conditional and depth diffusion models.

    3dgs inpainting scene-generation text-to-3d

    Project · GitHub · arXiv

  • SC

    SceneCraft Discovered by AI

    2024

    A layout-guided 3D scene generation framework that generates complex 3D scenes conditioned on free-form layouts and viewpoints, utilizing a diffusion-based pipeline to ensure spatial consistency and high-quality geometry.

    3d-scene diffusion layout-guided

    Project · GitHub · arXiv

  • ST

    SceneDreamer360: Text-Driven 3D-Consistent Scene Generation with Panoramic Gaussian Splatting Discovered by AI

    2024

    A text-driven 3D scene generation framework that leverages panoramic image generation and 3D Gaussian Splatting (3DGS) to synthesize high-quality, spatially consistent 3D environments with 360-degree views.

    3d-scene gaussian-splatting panorama text-to-3d

    Project · GitHub · arXiv

  • SC

    SceneTeller Discovered by AI

    2024

    A text-based 3D room design framework that generates high-quality indoor 3D scenes from natural language descriptions, facilitating interactive room planning and environment design.

    indoor-scene room-design text-to-3d

    GitHub · arXiv

  • SP

    SceneX: Procedural Controllable Large-scale Scene Generation Discovered by AI

    2024

    A procedural controllable large-scale scene generation framework that combines Large Language Models (LLMs) with procedural content generation (PCG) tools to automatically produce high-quality 3D natural scenes and urban environments from textual descriptions.

    3d-scene llm pcg procedural-generation

    Project · GitHub · arXiv

  • WO

    WonderWorld Discovered by AI

    2024

    An interactive 3D scene generation framework that allows users to progressively generate connected, diverse, and explorable 3D environments from a single image in low latency.

    3d-scene gaussian-splatting interactive

    GitHub · arXiv · Website

  • CR

    Ctrl-Room: Controllable Text-to-3D Room Meshes Generation with Layout Constraints Discovered by AI

    2023

    A controllable text-to-3D room generation framework that separates layout and appearance modeling, enabling the generation of high-fidelity 3D indoor scenes with designer-style layouts and supporting interactive editing operations like resizing or moving furniture.

    3d-scene indoor-scene layout-control text-to-3d

    Project · GitHub · arXiv

  • HO

    Holodeck Discovered by AI

    2023

    A system that generates diverse, interactive, and semantically rich 3D environments from natural language prompts by leveraging large language models for spatial planning and asset selection.

    3d-scene llm procedural

    GitHub · arXiv · Website

  • SC

    SceneDreamer Discovered by AI

    2023

    An unconditional generative model for unbounded 3D scenes that synthesizes large-scale 3D landscapes from random noise, trained solely on in-the-wild 2D image collections without 3D annotations.

    generative-model landscape-generation unbounded-scene

    GitHub · arXiv

Proprietary / Industry Tools

  • WL

    World Labs Marble Discovered by AI

    2025

    A multimodal 3D world model that generates persistent, high-fidelity 3D environments from text, images, videos, or coarse layouts. It supports interactive camera control, AI-native world editing (adding/removing objects, style changes), and exporting scenes as Gaussian splats, meshes, or video.

    3d-scene environment gaussian-splatting world-model

    Website

  • AA

    Atlas AI Studio Discovered by AI

    2023

    A generative AI platform designed for game developers to accelerate virtual worldbuilding by transforming 2D concept art, mood boards, and text prompts into structured 3D assets and environments.

    3d-assets concept-to-3d worldbuilding

    Website

  • PA

    Promethean AI Discovered by AI

    2019

    An AI-driven suite for virtual world building and set dressing that assists artists by automatically placing assets, suggesting layouts, and generating 3D environments within popular game engines.

    set-dressing unreal-engine world-building

    Website

  • GG

    Genie (Google DeepMind)

    DeepMind's foundation world model that generates interactive, playable 3D environments from a single image or text prompt, controllable in real time.

    commercial interactive playable world-model

    Website

  • MW

    Marble (World Labs)

    World Labs' commercial platform that generates large, explorable 3D worlds from a prompt or image, aimed at environment creation for games and virtual production.

    api commercial scene world-model

    Website

  • OD

    Odyssey

    An emerging commercial generative world model producing interactive, explorable environments for real-time world creation.

    commercial interactive world-model

    Website

  • SP

    Spaitial

    An emerging commercial spatial-intelligence platform for generating navigable 3D scenes and worlds from prompts or images.

    commercial scene world-model

    Website