From Visual Synthesis to Interactive Worlds: Toward Production-Ready 3D Asset Generation
Jiafeng Wu 1,2*β Β Β·Β
Zhuofan Lou 1,3*β Β Β·Β
Jian Liu 1β‘
Chunchao Guo 4 Β Β·Β
Dazhao Du 1 Β Β·Β
Song Guo 1Β§
1 The Hong Kong University of Science and Technology
2 Huazhong University of Science and Technology Β Β·Β
3 Sichuan University Β Β·Β
4 Tencent
* Equal contribution Β Β·Β β Work done during internship at HKUST Β Β·Β β‘ Project lead Β Β·Β Β§ Corresponding author
β¦ Production-Oriented 3D Generation Survey β¦
π§ Taxonomy Β·
π Data Β·
π¦ Objects Β·
π§ Characters Β·
π Scenes Β·
π Evaluation Β·
π Industry
π₯ [2026-04] v2.0: Introduces a production-ready two-dimensional taxonomy spanning asset tiers and pipeline stages, and expands the collection to data foundations, geometry, topology, appearance, rigging, scene assembly, evaluation, and industry systems.
π [2025-08] v1.0: Established the initial curated 3D AIGC paper collection, organized object, scene, and avatar methods under 3D-native, 2D-prior, and hybrid paradigms, and tracked surveys, datasets, talks, companies, and implementations.
Three-dimensional content generation has progressed from producing isolated, visually plausible shapes to constructing structured assets that can be deployed in real-time interactive environments. This trajectory is driven by converging demands from game development, embodied AI, world simulation, digital twins, and spatial computing, all of which require 3D content that goes beyond surface appearance to satisfy engine-level constraints on topology, UV parameterization, physically based materials, skeletal rigging, and physics-aware scene layout. Despite rapid advances in generative modeling, a persistent gap separates the outputs of current methods from the production-ready standard expected by interactive applications. This survey addresses that gap by organizing the literature around the asset production pipeline rather than algorithmic families.
At a Glance
π§ Organized by a production-ready, pipeline-first taxonomy rather than isolated algorithm families.
π¦ Covers three asset tiers: general objects, characters and avatars, and scenes and environments.
π Tracks the full asset workflow from data foundations through geometry, topology, appearance, rigging, and scene assembly.
π Consolidates methods, datasets, evaluation criteria, and industry references in one companion list.
The survey is organized around a two-dimensional taxonomy :
Horizontal axis (asset types): General Objects, Characters & Avatars, Scenes & Environments
Vertical axis (pipeline stages): Data Foundations β Geometry β Topology β Appearance β Rigging β Scene Assembly
This structure mirrors the production pipeline used in game engines and interactive applications, enabling direct assessment of where each method fits within a deployment workflow.
Data Foundations & Benchmarks
Dataset
Year
Scale
Description
ShapeNet
2015
51K models, 55 categories
Large-scale 3D shape repository (Chang et al.)
ModelNet
2015
12K CAD models, 40 categories
Princeton 3D object benchmark (Wu et al.)
ABC
2019
1M+ CAD models
Mechanical parts with parametric annotations (Koch et al.)
Thingi10K
2016
10K printable models
Web-derived 3D printing models with diverse topology (Zhou and Jacobson)
PartNet
2019
27K objects, 573K parts
Part-level object annotations for structural decomposition (Mo et al.)
Text2Shape
2018
75K text-shape pairs
Paired text and shape corpus for language-conditioned 3D generation (Chen et al.)
GSO (Google Scanned Objects)
2022
1K+ scans
Household objects with PBR materials (Downs et al.)
ABO (Amazon Berkeley Objects)
2022
8K+ models
Product catalog with multi-view images (Collins et al.)
CO3D
2021
1.5M frames, 19K objects
Multi-view real-capture object dataset for category-level reconstruction (Reizenstein et al.)
Objaverse
2023
800K+ objects
Internet-scale 3D asset collection (Deitke et al.)
Objaverse-XL
2024
10.2M objects
Extended internet-scale collection (Deitke et al.)
Dataset
Year
Scale
Description
FAUST
2014
300 scans, 10 subjects
Real body scans with ground-truth correspondence (Bogo et al.)
RenderPeople
2018
4.5K+ subjects
Commercially scanned textured human meshes for character production (RenderPeople)
AMASS
2019
Large-scale motion capture
Unified motion capture archive (Mahmood et al.)
CAPE
2020
4D clothing
Clothed body scans with pose variation (Ma et al.)
THuman2.0
2021
526 high-res scans
Detailed textured human models (Yu et al.)
HuMMan
2022
1K subjects
Multi-modal human dataset (Cai et al.)
HumanML3D
2022
14.6K text-motion sequences
Text-aligned motion corpus for controllable human generation and evaluation (Guo et al.)
Motion-X
2024
Large-scale motion
Expressive whole-body motion dataset (Lin et al.)
Dataset
Year
Scale
Description
ScanNet
2017
1,513 indoor scans
RGB-D reconstructions with annotations (Dai et al.)
ScanNet++
2023
460 scenes
Laser+DSLR indoor scans with material annotations (Yeshwanth et al.)
Matterport3D
2017
90 buildings
Large-scale indoor environments (Chang et al.)
HM3D
2021
1K buildings
Habitat-scale indoor scans with navigation annotations (Ramakrishnan et al.)
Structured3D
2020
3.5K houses
Synthetic indoor scenes with layout and topology labels (Zheng et al.)
Hypersim
2021
461 scenes, 77K images
Photorealistic synthetic scenes with material labels (Roberts et al.)
3D-FRONT
2021
18K rooms
Professionally designed indoor layouts (Fu et al.)
ProcTHOR
2022
Procedural houses
Infinitely scalable simulated interiors (Deitke et al.)
Infinigen
2023
Procedural nature
Photorealistic procedural generation of natural worlds (Raistrick et al.)
Infinigen Indoors
2024
Procedural interiors
Indoor extension of Infinigen (Raistrick et al.)
Method
Year
Venue
Highlight
π
π»
π
π
DreamFusion
2022
ICLR 2023
Pioneering open-domain text-to-3D via SDS
π
-
π
π
Magic3D
2023
CVPR 2023
Coarse-to-fine SDS for higher-resolution detail
π
-
π
π
Fantasia3D
2023
ICCV 2023
Disentangled geometry-appearance SDS with DMTet
π
π»
-
π
ProlificDreamer
2023
NeurIPS 2023
Variational SDS reducing over-smoothing
π
π»
-
π
RichDreamer
2024
CVPR 2024
Normal-depth diffusion prior for stable geometry
π
π»
-
π
Multi-View Reconstruction (MV)
Method
Year
Venue
Highlight
π
π»
π
π
Zero-1-to-3
2023
ICCV 2023
View-conditioned diffusion for novel-view synthesis
π
π»
-
π
MVDream
2024
ICML 2024
Multi-view consistent diffusion model
π
π»
-
π
Wonder3D
2024
CVPR 2024
Color + normal diffusion for normal-guided recon.
π
π»
-
π
SV3D
2024
ECCV 2024
Video diffusion for dense multi-view generation
π
-
π
π
Method
Year
Venue
Highlight
π
π»
π
π
3D-GAN
2016
NeurIPS 2016
Pioneering voxel-based adversarial 3D generation
π
-
-
π
Tree-GAN
2019
arXiv 2019
Tree-structured generator for point clouds
π
-
-
π
SP-GAN
2021
ICCV 2021
Spherical prior for global shape consistency
π
-
-
π
SDF-StyleGAN
2022
CVPR 2022
StyleGAN adapted for high-resolution SDF fields
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
AtlasNet
2018
CVPR 2018
Patch-deformation VAE for surface reconstruction
π
π»
-
π
TM-Net
2021
arXiv 2021
Joint geometry-texture VAE generation
π
-
-
π
Michelangelo
2023
NeurIPS 2023
Aligned shape-conditioned VAE with multimodal input
π
π»
-
π
CLAY
2024
arXiv 2024
Large-scale VAE + DiT for controllable 3D generation
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
PC-DPM
2021
ICLR 2021
DDPM for point cloud denoising generation
π
-
-
π
MeshDiffusion
2023
ICLR 2023
Score-based diffusion directly on mesh vertices
π
π»
-
π
TetraDiffusion
2024
arXiv 2024
Tetrahedral diffusion for high-resolution topology
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
Pixel2Mesh
2018
ECCV 2018
GCN-based mesh deformation from single image
π
π»
-
π
LRM
2023
ICLR 2024
Transformer-based large reconstruction model
π
-
π
π
TripoSR
2024
arXiv 2024
Distilled feed-forward for sub-second reconstruction
π
π»
-
π
GS-LRM
2024
ECCV 2024
Large reconstruction model for 3D Gaussian Splatting
π
-
π
π
InstantMesh
2024
arXiv 2024
Multi-view to FlexiCubes mesh with UV
π
π»
-
π
LGM
2024
arXiv 2024
Large multi-view Gaussian model for high-resolution 3D content
π
π»
-
π
SF3D
2024
arXiv 2024
Joint mesh + UV + PBR material prediction
π
π»
-
π
Fast3R
2025
arXiv 2025
Amortized scalable multi-view 3D reconstruction
π
-
-
π
Latent Generative Models (LGM)
Method
Year
Venue
Highlight
π
π»
π
π
Shap-E
2023
arXiv 2023
Fast text/image-to-3D via latent diffusion
π
π»
-
π
3DShape2VecSet
2023
SIGGRAPH 2023
Unordered vector-set VAE + diffusion
π
π»
-
π
LATTICE
2025
arXiv 2025
High-fidelity 3D generation at scale in compact latent space
π
-
-
π
XCube
2024
CVPR 2024
Hierarchical sparse-voxel latent diffusion
π
-
-
π
TRELLIS
2025
CVPR 2025
Structured latent (SLAT) with rectified flow
π
π»
-
π
TRELLIS.2
2025
arXiv 2025
O-Voxel representation, 4B params, PBR output
π
π»
-
π
SparseFlex
2025
arXiv 2025
Sparse isosurface VAE + flow for arbitrary topology
π
-
-
π
TripoSG
2025
arXiv 2025
VAE + rectified flow DiT for high-fidelity meshes
π
π»
-
π
MeshCraft
2025
arXiv 2025
Face-token VAE + flow DiT for parallel mesh gen.
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
PAGENet
2020
AAAI 2020
Part-aware generation network
-
-
-
π
SAMPart3D
2024
arXiv 2024
Multi-granularity zero-shot 3D part segmentation
-
-
-
π
HoloPart
2025
arXiv 2025
Amodal 3D part completion
-
-
-
π
X-Part
2025
arXiv 2025
Structure-coherent controllable shape decomposition
π
-
-
π
PartGen
2025
arXiv 2025
Part-level multi-view diffusion generation
-
-
-
π
PartCrafter
2025
arXiv 2025
Part-wise 3D reconstruction and editing
-
-
-
π
OmniPart
2025
arXiv 2025
Unified part-aware reconstruction pipeline
-
-
-
π
Indirect Methods (Post-hoc Remeshing)
Method
Year
Venue
Highlight
π
π»
π
π
Instant Meshes
2015
ACM TOG 2015
Field-aligned instant quad/tri remeshing
-
π»
-
π
QuadriFlow
2018
SGP 2018
Scalable instant field-aligned quad remeshing
π
π»
-
π
NeurCross
2025
arXiv 2025
Neural-guided cross-field remeshing
-
-
-
π
Direct Methods -- Autoregressive
Method
Year
Venue
Highlight
π
π»
π
π
PolyGen
2020
ICML 2020
Pioneering vertex-then-face autoregressive mesh gen.
π
π»
-
π
MeshGPT
2024
ICLR 2024
VQ-VAE codebook tokenization for mesh generation
π
-
π
π
MeshAnything
2024
arXiv 2024
Shape-conditioned artist mesh extraction
π
π»
-
π
MeshAnything V2
2025
arXiv 2025
Adjacent mesh tokenization with improved compression
π
π»
-
π
Meshtron
2024
arXiv 2024
Artist-like mesh generation at scale from artist-created data
π
-
-
π
PivotMesh
2024
arXiv 2024
Coarse-to-fine mesh scaffolding
π
π»
-
π
EdgeRunner
2024
arXiv 2024
Hybrid AR-latent mesh generation pipeline
π
π»
-
π
QuadGPT
2025
arXiv 2025
Autoregressive quad-mesh generation for edge loops
π
-
-
π
DeepMesh
2025
ICCV 2025
RL-based preference alignment for artist-quality mesh
π
π»
-
π
Mesh-RFT
2025
arXiv 2025
Masked DPO for localized mesh defect correction
π
-
-
π
Direct Methods -- Diffusion
Method
Year
Venue
Highlight
π
π»
π
π
PolyDiff
2023
ICCV 2023
Discrete denoising diffusion over triangle soups
π
-
-
π
SpaceMesh
2024
arXiv 2024
Continuous halfedge latent diffusion; ultra-fast
π
-
-
π
MeshCraft
2025
arXiv 2025
Flow-based DiT with face-count control
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
xatlas
2018
Open-source
Production-baseline automatic UV atlas packer
-
π»
-
π
UVAtlas
2023
Open-source
Production-baseline atlas generation for UV layout
-
π»
-
π
Auto-UV
2025
arXiv 2025
Learned seam prediction for UV unwrapping
-
-
-
π
Flatten Anything
2024
arXiv 2024
Unsupervised cycle-consistent UV mapping
-
-
-
π
FlexPara
2025
arXiv 2025
Flexible unsupervised UV parameterization
π
-
-
π
PartUV
2025
arXiv 2025
Semantic chart-aligned UV partitioning
-
-
-
π
ArtUV
2025
arXiv 2025
Artist-style UV packing and layout
π
-
-
π
SeamCrafter
2025
arXiv 2025
Seam preference optimization for UV quality
π
-
-
π
Hunyuan3D Studio
2025
arXiv 2025
End-to-end game-ready asset pipeline with integrated UV construction
π
-
-
π
Texture & PBR Material Generation
Method
Year
Venue
Highlight
π
π»
π
π
TEXTure
2023
SIGGRAPH 2023
Iterative text-guided texture painting
π
π»
-
π
Text2Tex
2023
ICCV 2023
Progressive inpainting for mesh texturing
π
π»
-
π
TexFusion
2023
arXiv 2023
Cross-view aggregated texture fusion
π
-
-
π
Paint3D
2024
arXiv 2024
Illumination-free texture diffusion for PBR
π
π»
-
π
FlashTex
2024
arXiv 2024
Fast text-to-texture generation
π
-
-
π
TexGen
2024
arXiv 2024
Feed-forward UV-space texture diffusion
-
-
-
π
MVPaint
2025
arXiv 2025
Multi-view consistent texture painting
π
-
-
π
MaterialMVP
2025
ICCV 2025
Illumination-invariant multi-view PBR diffusion
-
-
-
π
MaterialAnything
2024
arXiv 2024
PBR material decomposition as first-class objective
π
-
-
π
3DTopia-XL
2024
arXiv 2024
Primitive diffusion for high-quality 3D assets with joint UV/PBR objectives
π
π»
-
π
Meta 3D AssetGen
2024
arXiv 2024
Unified UV + geometry + PBR material pipeline
π
-
-
π
PBR3DGen
2025
arXiv 2025
VLM-guided mesh generation with PBR materials
π
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
SMPL
2015
SIGGRAPH Asia 2015
Skinned multi-person linear body model
π
-
-
π
SMPL-X
2019
CVPR 2019
Expressive body + hands + face parametric model
-
-
-
π
FLAME
2017
SIGGRAPH Asia 2017
Learned head model with expression blendshapes
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
Tex2Shape
2019
ICCV 2019
UV-space displacement prediction on SMPL
-
-
-
π
CAPE
2020
CVPR 2020
Pose-dependent clothing offsets on body mesh
-
-
-
π
ExPose
2020
ECCV 2020
Monocular body + hands + face estimation
-
-
-
π
STAR
2020
ECCV 2020
Sparse parametric SMPL variant
-
-
-
π
HybrIK
2021
CVPR 2021
Hybrid analytical-regressive IK for body mesh
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
PIFu
2019
ICCV 2019
Pixel-aligned implicit function for clothed humans
-
-
-
π
ARCH
2020
CVPR 2020
Canonical implicit field with SMPL-guided rigging
-
-
-
π
PIFuHD
2020
CVPR 2020
Multi-level implicit for high-res human recon.
-
-
-
π
PaMIR
2021
TPAMI 2021
Parametric body model inside implicit recon.
-
-
-
π
SMPLicit
2021
CVPR 2021
Implicit clothing conditioned on SMPL parameters
π
π»
-
π
ICON
2022
CVPR 2022
Normal-guided implicit body with SMPL
-
-
-
π
gDNA
2022
ECCV 2022
Generative implicit model for diverse humans
-
-
-
π
ECON
2023
CVPR 2023
Implicit + explicit mesh hybrid reconstruction
-
-
-
π
S3F
2023
ICCV 2023
Structured 3D features with SMPL semi-supervision
-
-
-
π
GAN / Diffusion Generative
Method
Year
Venue
Highlight
π
π»
π
π
StylePeople
2021
arXiv 2021
GAN-based clothed mesh synthesis
-
-
-
π
AvatarGen
2022
arXiv 2022
SDF + tri-plane GAN for 3D avatars
-
-
-
π
AvatarCLIP
2022
SIGGRAPH 2022
CLIP-guided text-to-avatar generation
π
π»
-
π
Get3DHuman
2023
arXiv 2023
Tri-plane/SDF GAN for full-body humans
π
π»
-
π
GETAvatar
2023
NeurIPS 2023
GAN + SMPL for animatable textured mesh
-
-
-
π
AvatarCraft
2023
arXiv 2023
SDS-based NeRF-to-mesh avatar generation
-
-
-
π
DreamHuman
2023
arXiv 2023
SDS + imGHUM body prior for text-to-human
π
-
-
π
ChuPa
2023
arXiv 2023
2D diffusion on SMPL with displacement
-
-
-
π
DreamAvatar
2024
arXiv 2024
SDS + SMPL-guided NeRF avatar synthesis
-
-
-
π
Morphable Diffusion
2024
CVPR 2024
3D-consistent diffusion for single-image avatar creation
π
-
-
π
SiTH
2024
CVPR 2024
Single-view textured human reconstruction with image-conditioned diffusion
π
-
-
π
TADA!
2024
CVPR 2024
SMPL-X + texture with LBS-ready rigging
π
π»
-
π
Method
Year
Venue
Highlight
π
π»
π
π
InstantAvatar
2023
CVPR 2023
Fast neural field from monocular video
-
-
-
π
SHERF
2023
arXiv 2023
Canonical-prior NeRF for generalizable humans
-
-
-
π
LRM
2023
ICLR 2024
Feed-forward large reconstruction model adapted to human body synthesis
π
-
π
π
Human GS
2024
arXiv 2024
Canonical 3DGS with skinning from multi-view
-
-
-
π
HUGS
2024
CVPR 2024
3DGS + SMPL for real-time avatar playback
-
-
-
π
3DGS-Avatar
2024
arXiv 2024
Deformable 3DGS bound to skinning weights
π
π»
-
π
LHM
2025
arXiv 2025
Dense Gaussians on SMPL; animatable in seconds
-
-
-
π
OmniAvatar
2025
arXiv 2025
Video diffusion for temporally coherent avatars
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
i3DMM
2021
arXiv 2021
Implicit SDF-based 3D morphable model
-
-
-
π
NerFace
2021
arXiv 2021
NeRF with expression-conditioned deformation
-
-
-
π
EG3D
2022
CVPR 2022
Efficient tri-plane GAN for 3D face generation
-
-
-
π
NPHM
2023
arXiv 2023
Dual-SDF for identity + expression disentanglement
-
-
-
π
Next3D
2023
CVPR 2023
Tri-plane + texture rasterization with FLAME
-
-
-
π
PanoHead
2023
CVPR 2023
Depth-aware tri-grid for full 360-degree heads
-
-
-
π
RODIN
2023
arXiv 2023
Diffusion on tri-plane for novel-ID head generation
-
-
-
π
HeadSculpt
2023
arXiv 2023
SDS NeRF with SMPL-X prior for head sculpting
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
GaussianAvatars
2024
CVPR 2024
3DGS bound to FLAME triangles for animation
-
-
-
π
FlashAvatar
2024
arXiv 2024
>300 FPS 3DGS + FLAME; production-ready
-
-
-
π
MonoGaussianAvatar
2024
arXiv 2024
Monocular-video 3DGS on FLAME mesh
-
-
-
π
RGCA (Relightable)
2024
arXiv 2024
Full PBR 3DGS decomposition for relighting
-
-
-
π
Feed-Forward Reconstruction
Method
Year
Venue
Highlight
π
π»
π
π
GAGAvatar
2024
arXiv 2024
Generalizable 3DGS from single image with FLAME
-
-
-
π
Arc2Avatar
2025
arXiv 2025
Identity-guided single-image 3DGS head recon.
-
-
-
π
HRAvatar
2025
arXiv 2025
PBR 3DGS with roughness + Fresnel from mono video
-
-
-
π
LAM
2025
arXiv 2025
Large Avatar Model; 280 FPS, LBS-compatible
-
-
-
π
Avat3r
2025
arXiv 2025
Sparse-view canonical 3DGS via ViT
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
SadTalker
2023
CVPR 2023
Audio-to-FLAME coefficients for talking-head video
-
-
-
π
HunyuanVideo-Avatar
2025
arXiv 2025
High-fidelity audio-driven human animation for multiple characters
π
-
-
π
TexTalker
2025
arXiv 2025
Audio-sync wrinkle maps + geometric deformation
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
RigNet
2020
SIGGRAPH 2020
End-to-end GNN skeleton + skinning prediction
-
-
-
π
SkinningNet
2022
arXiv 2022
Two-stream GNN for heterogeneous skeletal topologies
-
-
-
π
DeePSD
2021
arXiv 2021
Unsupervised physics-based garment skinning
-
-
-
π
Note: TADA!, ChuPa, LAM, and HRAvatar also include rigging capabilities -- see above.
Method
Year
Venue
Highlight
π
π»
π
π
ATISS
2021
NeurIPS 2021
Autoregressive transformer for room layouts
π
π»
-
π
ProcTHOR
2022
NeurIPS 2022
Procedural interactive houses for embodied AI
-
π»
π
π
Pose2Room
2022
arXiv 2022
Activity-driven affordance-aware room layout
-
-
-
π
DiffuScene
2024
arXiv 2024
Diffusion + retrieval for furnished indoor scenes
π
π»
-
π
Holodeck
2024
CVPR 2024
LLM-planned embodied environment generation
-
-
-
π
LayoutGPT
2023
arXiv 2023
LLM-generated indoor layouts from text
-
-
-
π
LLplace
2024
arXiv 2024
Dialogue-driven interactive layout editing
π
-
-
π
CityCraft
2024
arXiv 2024
Language-guided city-scale layout generation
π
-
-
π
Scene Population & Asset Grounding
Method
Year
Venue
Highlight
π
π»
π
π
MIME
2023
arXiv 2023
Human-motion-informed object placement
-
-
-
π
AnyHome
2023
arXiv 2023
Open-vocabulary text-to-house scene generation
π
-
-
π
Open-Universe
2024
arXiv 2024
LLM programs + solver for open-vocabulary scenes
π
-
-
π
SceneCraft
2024
arXiv 2024
Blender code agent for executable scene scripts
-
-
-
π
PhyScene
2024
arXiv 2024
Physics-guided diffusion for interactable scenes
-
-
-
π
UnrealLLM
2025
arXiv 2025
Unreal Engine PCG agents from language
-
-
-
π
Layout2Scene
2025
arXiv 2025
Layout-guided diffusion for holistic scenes
π
-
-
π
3D-GPT
2023
arXiv 2023
Procedural modeling as language-conditioned programs
π
π»
-
π
PhysGen3D
2025
arXiv 2025
Miniature interactive worlds with physics simulation
-
-
-
π
Method
Year
Venue
Highlight
π
π»
π
π
Text2Light
2022
arXiv 2022
Text-conditioned HDR panorama for skybox/lighting
π
π»
-
π
Text2Room
2023
arXiv 2023
2D diffusion lifted to textured room meshes
π
π»
-
π
Infinigen
2023
CVPR 2023
Photorealistic procedural natural world generation
-
-
-
π
CityDreamer
2024
arXiv 2024
Unbounded urban synthesis with stuff + thing fields
π
π»
-
π
Infinigen Indoors
2024
arXiv 2024
Photorealistic procedural indoor worlds
-
-
-
π
LayerPano3D
2025
arXiv 2025
Layered panorama to explorable 3DGS scene
-
-
-
π
WorldCraft
2025
arXiv 2025
LLM-agentic world editing and customization
π
-
-
π
The survey identifies several complementary evaluation dimensions for production-ready 3D generation:
Dimension
Metrics
Geometric Fidelity
Chamfer Distance (CD), Earth Mover's Distance (EMD), F-Score, Normal Consistency, Coverage (COV), Minimum Matching Distance (MMD), 1-NNA
Appearance Quality
PSNR, SSIM, LPIPS, FID, KID, CLIP Score, CLIP R-Precision
PBR / Relighting
Relighting consistency, albedo/roughness/metallic separation quality, paper-specific material decomposition tests
Asset Usability
UV stretch and angular distortion, seam visibility, chart packing efficiency, overlap detection, rig smoothness, retargeting success, engine import success
Topology Readiness
Manifoldness, watertightness, genus correctness, quad ratio, edge-flow alignment, collision-mesh quality
Scene-Level
Physical plausibility, interpenetration, NavMesh connectivity, navigation success rate, affordance compatibility, A/B preference, Likert ratings
A key finding of this survey is that existing benchmarks systematically overestimate deployment readiness by focusing on geometric and appearance metrics while neglecting topology, UV/PBR, engine import, and other asset usability criteria required for interactive applications.
If you find this survey useful, please cite our paper:
@article {wu2026visual ,
title ={ From Visual Synthesis to Interactive Worlds: Toward Production-Ready 3D Asset Generation} ,
author ={ Wu, Jiafeng and Lou, Zhuofan and Liu, Jian and Du, Dazhao and Guo, Chunchao and Guo, Song} ,
journal ={ arXiv preprint arXiv:2604.23629} ,
year ={ 2026}
}
If you also use resources from the v1 collection, please additionally cite:
@article {liu2024comprehensive ,
title ={ A Comprehensive Survey on 3D Content Generation} ,
author ={ Liu, Jian and Huang, Xiaoshui and Huang, Tianyu and Chen, Lu and Hou, Yuenan and Tang, Shixiang and Liu, Ziwei and Ouyang, Wanli and Zuo, Wangmeng and Jiang, Junjun and others} ,
journal ={ arXiv preprint arXiv:2402.01166} ,
year ={ 2024}
}
We welcome contributions! Please see CONTRIBUTING.md for guidelines.
Note: Pull requests should target the v2 branch. The main branch preserves the original v1 awesome list.
This work was supported by The Hong Kong University of Science and Technology and Tencent.
The original awesome list curated is preserved on the main branch . It contains a broader collection of AIGC 3D papers organized by topic without the production-pipeline focus of v2.