arxiv: 2506.16504 · v1 · submitted 2025-06-19 · 💻 cs.CV · cs.AI

Recognition: 1 theorem link

Hunyuan3D 2.5: Towards High-Fidelity 3D Assets Generation with Ultimate Details

Zeqiang Lai , Yunfei Zhao , Haolin Liu , Zibo Zhao , Qingxiang Lin , Huiwen Shi , Xianghui Yang , Mingxin Yang

show 18 more authors

Shuhui Yang Yifei Feng Sheng Zhang Xin Huang Di Luo Fan Yang Fang Yang Lifu Wang Sicong Liu Yixuan Tang Yulin Cai Zebin He Tian Liu Yuhong Liu Jie Jiang Linus Jingwei Huang Chunchao Guo

Authors on Pith no claims yet

Pith reviewed 2026-05-16 14:05 UTC · model grok-4.3

classification 💻 cs.CV cs.AI

keywords 3D generationdiffusion modelsshape generationtexture generationPBR textureshigh-fidelity 3D assetsmodel scalingLATTICE

0 comments

The pith

Scaling a shape foundation model to 10 billion parameters yields sharp, detailed 3D meshes and PBR textures that closely match input images.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents Hunyuan3D 2.5, a two-stage diffusion system for turning images into high-fidelity textured 3D assets. It introduces LATTICE, a new shape model trained at much larger scale with 10 billion parameters, expanded high-quality data, and more compute. This produces meshes that follow the input image precisely while staying geometrically clean and smooth. Texture generation is upgraded with physical-based rendering through an extended multi-view architecture. The combined system outperforms earlier methods and narrows the quality difference between generated and handcrafted 3D objects.

Core claim

Hunyuan3D 2.5 introduces LATTICE, a shape foundation model scaled to 10B parameters with larger high-quality datasets and increased compute, which generates sharp and detailed 3D shapes with precise image-3D following while keeping mesh surfaces clean and smooth. Texture generation is upgraded with physical-based rendering via a novel multi-view architecture extended from the prior Paint model. The full system significantly outperforms previous methods in both shape generation and end-to-end texture quality.

What carries the argument

LATTICE, the scaled shape foundation model that uses expanded model size to 10B parameters, datasets, and compute to drive improvements in 3D mesh detail, alignment, and surface quality.

If this is right

The 10B-parameter model produces 3D shapes that are both highly detailed and free of surface artifacts.
Precise following of 2D images to 3D outputs is achieved without trading off geometric cleanliness.
PBR textures generated through the multi-view stage increase realism across different renderings.
The overall pipeline reduces the quality gap between automatically generated and handcrafted 3D assets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Scaling approaches that worked for 2D images may transfer to 3D asset creation with similar benefits.
Industries such as gaming and virtual reality could gain from faster production of realistic 3D content.
Tighter coupling between the shape and texture stages might reduce inconsistencies in final assets.

Load-bearing premise

That scaling model size, training data, and compute will directly deliver the claimed gains in shape fidelity and texture quality without overfitting or evaluation biases favoring the new system.

What would settle it

A quantitative benchmark or blind user study on standard 3D generation metrics where Hunyuan3D 2.5 shows no improvement or lower scores than prior methods in shape detail, surface smoothness, or texture accuracy.

read the original abstract

In this report, we present Hunyuan3D 2.5, a robust suite of 3D diffusion models aimed at generating high-fidelity and detailed textured 3D assets. Hunyuan3D 2.5 follows two-stages pipeline of its previous version Hunyuan3D 2.0, while demonstrating substantial advancements in both shape and texture generation. In terms of shape generation, we introduce a new shape foundation model -- LATTICE, which is trained with scaled high-quality datasets, model-size, and compute. Our largest model reaches 10B parameters and generates sharp and detailed 3D shape with precise image-3D following while keeping mesh surface clean and smooth, significantly closing the gap between generated and handcrafted 3D shapes. In terms of texture generation, it is upgraded with phyiscal-based rendering (PBR) via a novel multi-view architecture extended from Hunyuan3D 2.0 Paint model. Our extensive evaluation shows that Hunyuan3D 2.5 significantly outperforms previous methods in both shape and end-to-end texture generation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Hunyuan3D 2.5 scales the shape model to 10B parameters and adds a multi-view PBR texture stage on top of the prior 2.0 pipeline, but the abstract supplies no metrics or ablations to support the fidelity gains.

read the letter

The main point is that this work extends the authors' own two-stage diffusion pipeline by training a new 10B-parameter shape model called LATTICE on larger datasets and compute, then upgrading the texture stage with a multi-view physical-based rendering architecture. The shape part emphasizes clean meshes and tight image-3D alignment, while the texture upgrade targets better material consistency across views. Both build directly on Hunyuan3D 2.0 rather than introducing a new paradigm, so the novelty sits in the scale and the specific multi-view texture design. If the full paper shows the architecture details and training recipe clearly, that could be useful for groups already running similar two-stage setups. The central weakness is the lack of any numbers. The abstract asserts that the largest model closes the gap to handcrafted assets and significantly outperforms prior methods, yet it reports zero quantitative results—no Chamfer distances, normal consistency scores, IoU, user-study percentages, or ablations that isolate model size from data quality. Without those, it is impossible to tell whether the claimed improvements come from the extra parameters or from dataset curation and inference choices. The stress-test concern holds up on the provided text: the scaling story needs evidence that bigger is actually better here. This paper is mainly for researchers tracking industrial 3D generation progress in gaming and simulation who want to see the latest scaling attempt from this group. A reader already working on multi-view texture or large 3D diffusion models could extract the architectural choices, but the missing evaluation makes it hard to assess impact. I would send it to peer review if the full manuscript contains the promised quantitative comparisons and ablations; the subfield benefits from documented scaling results even when they are incremental.

Referee Report

1 major / 0 minor

Summary. The manuscript presents Hunyuan3D 2.5, a two-stage 3D diffusion model suite for high-fidelity textured 3D assets. Building on Hunyuan3D 2.0, it introduces the LATTICE shape foundation model scaled to 10B parameters via larger high-quality datasets and compute, claimed to produce sharp detailed shapes with precise image-3D alignment and clean smooth surfaces that close the gap to handcrafted meshes. The texture stage is upgraded to physical-based rendering (PBR) using a novel multi-view architecture. The paper asserts that extensive evaluations show significant outperformance over prior methods in both shape and end-to-end texture generation.

Significance. If substantiated by rigorous quantitative comparisons, the scaling of LATTICE to 10B parameters and the PBR texture upgrade could mark a notable advance in closing the quality gap between generated and professional 3D assets, demonstrating the value of large-scale training for 3D fidelity. The work would strengthen evidence that model size, data, and compute scaling translate to measurable gains in sharpness, alignment, and surface quality for downstream applications in graphics and content creation.

major comments (1)

[Abstract] Abstract: The central claim that the 10B-parameter LATTICE model 'significantly outperforms previous methods' and 'significantly closing the gap between generated and handcrafted 3D shapes' is unsupported by any quantitative metrics, baselines, ablation studies, or error analysis (e.g., no Chamfer distance, IoU, normal consistency, or user-study scores). Without these, it is impossible to isolate the contribution of model scaling from dataset curation or inference choices, rendering the scaling hypothesis unevaluable.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback on our manuscript. We agree that the abstract's claims require stronger quantitative backing to allow readers to evaluate the scaling hypothesis for LATTICE. We will revise the paper to include explicit metrics, baselines, and ablations while preserving the core technical contributions.

read point-by-point responses

Referee: [Abstract] Abstract: The central claim that the 10B-parameter LATTICE model 'significantly outperforms previous methods' and 'significantly closing the gap between generated and handcrafted 3D shapes' is unsupported by any quantitative metrics, baselines, ablation studies, or error analysis (e.g., no Chamfer distance, IoU, normal consistency, or user-study scores). Without these, it is impossible to isolate the contribution of model scaling from dataset curation or inference choices, rendering the scaling hypothesis unevaluable.

Authors: We accept this critique. The current abstract summarizes results without citing specific numbers, and the experiments section relies primarily on qualitative comparisons and visual results rather than tabulated metrics such as Chamfer distance, IoU, normal consistency, or user-study scores. This makes it difficult to isolate the effect of scaling to 10B parameters. In the revision we will (1) add a quantitative comparison table reporting Chamfer distance, IoU, normal consistency, and user-study scores against prior methods, (2) include an ablation study on model size, data scale, and compute, and (3) revise the abstract to reference these concrete results. These additions will make the scaling hypothesis directly evaluable. revision: yes

Circularity Check

0 steps flagged

No significant circularity; claims rest on empirical scaling and external comparisons

full rationale

The paper describes an empirical 3D diffusion model (LATTICE) trained at scale, with performance claims tied to larger model size, datasets, and compute, followed by 'extensive evaluation' showing outperformance versus prior methods. No derivation chain exists that reduces outputs to self-defined inputs, fitted parameters renamed as predictions, or load-bearing self-citations whose validity depends on the current work. References to Hunyuan3D 2.0 describe architectural continuity but do not substitute for the reported gains, which remain falsifiable via independent benchmarks. This matches the default case of a non-circular empirical report.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

Central claim depends on the assumption that larger scale in data, model size, and compute yields superior 3D fidelity; limited details available from abstract only.

free parameters (1)

LATTICE model size
Scaled to 10B parameters to achieve claimed performance gains.

axioms (1)

domain assumption Scaling laws for 3D diffusion models hold and produce better image-3D alignment and mesh quality
Invoked when stating that increased model size, data, and compute close the gap to handcrafted shapes.

invented entities (1)

LATTICE no independent evidence
purpose: Shape foundation model for high-fidelity 3D generation
New named model introduced to deliver the claimed sharp and clean outputs.

pith-pipeline@v0.9.0 · 5587 in / 1413 out tokens · 36189 ms · 2026-05-16T14:05:48.164751+00:00 · methodology

discussion (0)

Forward citations

Cited by 20 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

R-DMesh: Video-Guided 3D Animation via Rectified Dynamic Mesh Flow
cs.CV 2026-05 unverdicted novelty 7.0

R-DMesh generates high-fidelity 4D meshes aligned to video by disentangling base mesh, motion, and a learned rectification jump offset inside a VAE, then using Triflow Attention and rectified-flow diffusion.
Velocity-Space 3D Asset Editing
cs.GR 2026-05 unverdicted novelty 7.0

VS3D performs local 3D asset editing by injecting reconstruction-anchored source signals, partial-mean guidance, and twin-agreement residuals into the velocity sampler to control edit strength and preserve identity.
THOM: Generating Physically Plausible Hand-Object Meshes From Text
cs.CV 2026-04 unverdicted novelty 7.0

THOM is a training-free two-stage framework that generates physically plausible hand-object 3D meshes directly from text by combining text-guided Gaussians with contact-aware physics optimization and VLM refinement.
ATATA: One Algorithm to Align Them All
cs.CV 2026-01 unverdicted novelty 7.0

ATATA enables fast joint inference of structurally aligned pairs using Rectified Flow models via segment transport, improving state-of-the-art for image and video generation while matching 3D quality at much higher speed.
Curvature-Aware Captioning:Leveraging Geodesic Attention for 3D Scene Understanding
cs.CV 2026-05 unverdicted novelty 6.0

A new framework combines self-attention on the Oblique manifold with bidirectional geodesic cross-attention on the Lorentz hyperboloid to improve both localization accuracy and descriptive coherence in 3D dense captioning.
DVD: Discrete Voxel Diffusion for 3D Generation and Editing
cs.CV 2026-05 unverdicted novelty 6.0

DVD treats voxel occupancy as a discrete variable in a diffusion framework to generate, assess, and edit sparse 3D voxels without continuous thresholding.
Toward Visually Realistic Simulation: A Benchmark for Evaluating Robot Manipulation in Simulation
cs.RO 2026-05 unverdicted novelty 6.0

VISER is a new visually realistic simulation benchmark for robot manipulation tasks that uses PBR materials and MLLM-assisted asset generation, achieving 0.92 Pearson correlation with real-world policy performance.
High-Fidelity Single-Image Head Modeling with Industry-Grade Topology
cs.CV 2026-05 unverdicted novelty 6.0

A single-image head reconstruction method uses coarse-to-fine optimization with normal consistency, landmarks, and geometry-aware constraints on curvature and conformality to produce meshes with industry-grade topolog...
Animator-Centric Skeleton Generation on Objects with Fine-Grained Details
cs.GR 2026-04 unverdicted novelty 6.0

An animator-centric skeleton generation method that uses semantic-aware tokenization and a learnable density interval module to produce controllable, high-quality skeletons on complex 3D meshes.
Beyond Voxel 3D Editing: Learning from 3D Masks and Self-Constructed Data
cs.CV 2026-04 unverdicted novelty 6.0

BVE framework enables text-guided 3D editing beyond voxel limits by combining self-constructed data, lightweight semantic injection, and annotation-free masking to preserve local invariance.
Grasp in Gaussians: Fast Monocular Reconstruction of Dynamic Hand-Object Interactions
cs.CV 2026-04 unverdicted novelty 6.0

GraG reconstructs dynamic 3D hand-object interactions from monocular video 6.4x faster than prior work by using compact Sum-of-Gaussians tracking initialized from large models and refined with 2D losses.
Pair2Scene: Learning Local Object Relations for Procedural Scene Generation
cs.CV 2026-04 unverdicted novelty 6.0

Pair2Scene generates complex 3D scenes beyond training data by training a network on local object-pair placement rules and applying them recursively with collision-aware sampling.
Pair2Scene: Learning Local Object Relations for Procedural Scene Generation
cs.CV 2026-04 unverdicted novelty 6.0

Pair2Scene generates complex 3D scenes beyond training data by recursively applying a learned model of local support and functional object-pair relations inside hierarchies, using collision-aware rejection sampling fo...
DailyArt: Discovering Articulation from Single Static Images via Latent Dynamics
cs.CV 2026-04 unverdicted novelty 6.0

DailyArt recovers full joint parameters of articulated objects from a single static image by synthesizing an opened state and comparing discrepancies, supporting downstream part-level novel state synthesis.
UniRecGen: Unifying Multi-View 3D Reconstruction and Generation
cs.CV 2026-04 unverdicted novelty 6.0

UniRecGen unifies reconstruction and generation via shared canonical space and disentangled cooperative learning to produce complete, consistent 3D models from sparse views.
StoryBlender: Inter-Shot Consistent and Editable 3D Storyboard with Spatial-temporal Dynamics
cs.CV 2026-04 unverdicted novelty 6.0

StoryBlender generates inter-shot consistent editable 3D storyboards using a three-stage pipeline of semantic-spatial grounding, canonical asset materialization, and spatial-temporal dynamics with agent-based verification.
R-DMesh: Video-Guided 3D Animation via Rectified Dynamic Mesh Flow
cs.CV 2026-05 unverdicted novelty 5.0

R-DMesh uses a VAE with a learned rectification jump offset and Triflow Attention inside a rectified-flow diffusion transformer to produce video-aligned 4D meshes despite initial pose misalignment.
DataEvolver: Let Your Data Build and Improve Itself via Goal-Driven Loop Agents
cs.AI 2026-05 unverdicted novelty 5.0

DataEvolver introduces a reusable framework with generation-time self-correction and validation-time self-expansion loops that improves visual datasets, shown to outperform baselines on an object-rotation task.
Asset Harvester: Extracting 3D Assets from Autonomous Driving Logs for Simulation
cs.CV 2026-04 unverdicted novelty 5.0

Asset Harvester converts sparse in-the-wild object observations from AV driving logs into complete simulation-ready 3D assets via data curation, geometry-aware preprocessing, and a SparseViewDiT model that couples spa...
Hitem3D 2.0: Multi-View Guided Native 3D Texture Generation
cs.CV 2026-04 unverdicted novelty 5.0

Hitem3D 2.0 combines multi-view image synthesis with native 3D texture projection to improve completeness, cross-view consistency, and geometry alignment over prior methods.

Reference graph

Works this paper leans on

24 extracted references · 24 canonical work pages · cited by 18 Pith papers · 4 internal anchors

[1]

Matatlas: Text-driven consistent geometry texturing and material assignment

Duygu Ceylan, Valentin Deschaintre, Thibault Groueix, Rosalie Martin, Chun-Hao Huang, Ro- main Rouffet, Vladimir Kim, and Ga¨etan Lassagne. Matatlas: Text-driven consistent geometry texturing and material assignment. arXiv preprint arXiv:2404.02899,

work page arXiv
[2]

Text2tex: Text-driven texture synthesis via diffusion models

Dave Zhenyu Chen, Yawar Siddiqui, Hsin-Ying Lee, Sergey Tulyakov, and Matthias Nießner. Text2tex: Text-driven texture synthesis via diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pp. 18558–18568, 2023a. Rui Chen, Yongwei Chen, Ningxin Jiao, and Kui Jia. Fantasia3d: Disentangling geometry and appearance for hi...

work page arXiv
[3]

Chenjian Gao, Boyan Jiang, Xinghui Li, Yingpeng Zhang, and Qian Yu

URL https://arxiv.org/abs/2503.19011. Chenjian Gao, Boyan Jiang, Xinghui Li, Yingpeng Zhang, and Qian Yu. Genesistex: adapting image denoising diffusion to texture space. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pp. 4620–4629,

work page arXiv
[4]

Meshtron: High-fidelity, artist-like 3d mesh generation at scale

Zekun Hao, David W Romero, Tsung-Yi Lin, and Ming-Yu Liu. Meshtron: High-fidelity, artist-like 3d mesh generation at scale. arXiv preprint arXiv:2412.09548,

work page arXiv
[5]

LRM: Large Reconstruction Model for Single Image to 3D

Yicong Hong, Kai Zhang, Jiuxiang Gu, Sai Bi, Yang Zhou, Difan Liu, Feng Liu, Kalyan Sunkavalli, Trung Bui, and Hao Tan. Lrm: Large reconstruction model for single image to 3d. arXiv preprint arXiv:2311.04400,

work page internal anchor Pith review Pith/arXiv arXiv
[6]

Material anything: Generating materials for any 3d object via diffusion

Xin Huang, Tengfei Wang, Ziwei Liu, and Qing Wang. Material anything: Generating materials for any 3d object via diffusion. arXiv preprint arXiv:2411.15138, 2024a. Zehuan Huang, Yuanchen Guo, Haoran Wang, Ran Yi, Lizhuang Ma, Yan-Pei Cao, and Lu Sheng. Mv-adapter: Multi-view consistent image generation made easy. arXiv preprint arXiv:2412.03632, 2024b. Te...

work page arXiv
[7]

Diederik P Kingma

URL https://arxiv.org/abs/2506.15442. Diederik P Kingma. Auto-encoding variational bayes. arXiv preprint arXiv:1312.6114,

work page arXiv
[8]

Peng Li, Yuan Liu, Xiaoxiao Long, Feihu Zhang, Cheng Lin, Mengfei Li, Xingqun Qi, Shanghang Zhang, Wenhan Luo, Ping Tan, et al

URL https://arxiv.org/abs/2503.16302. Peng Li, Yuan Liu, Xiaoxiao Long, Feihu Zhang, Cheng Lin, Mengfei Li, Xingqun Qi, Shanghang Zhang, Wenhan Luo, Ping Tan, et al. Era3d: High-resolution multiview diffusion using efficient row-wise attention. arXiv preprint arXiv:2405.11616, 2024a. Weiyu Li, Jiarui Liu, Hongyu Yan, Rui Chen, Yixun Liang, Xuelin Chen, Pi...

work page arXiv
[9]

Text-guided texturing by synchronized multi-view diffusion

Yuxin Liu, Minshan Xie, Hanyuan Liu, and Tien-Tsin Wong. Text-guided texturing by synchronized multi-view diffusion. In SIGGRAPH Asia 2024 Conference Papers , pp. 1–11, 2024a. Zexiang Liu, Yangguang Li, Youtian Lin, Xin Yu, Sida Peng, Yan-Pei Cao, Xiaojuan Qi, Xiaoshui Huang, Ding Liang, and Wanli Ouyang. Unidream: Unifying diffusion priors for relightabl...

work page 2024
[10]

Texture: Text-guided texturing of 3d shapes

Elad Richardson, Gal Metzer, Yuval Alaluf, Raja Giryes, and Daniel Cohen-Or. Texture: Text-guided texturing of 3d shapes. In ACM SIGGRAPH 2023 conference proceedings , pp. 1–11,

work page 2023
[11]

Matfusion: a generative diffusion model for svbrdf capture

Sam Sartor and Pieter Peers. Matfusion: a generative diffusion model for svbrdf capture. In SIGGRAPH Asia 2023 Conference Papers , pp. 1–10,

work page 2023
[12]

Zero123++: a Single Image to Consistent Multi-view Diffusion Base Model

Ruoxi Shi, Hansheng Chen, Zhuoyang Zhang, Minghua Liu, Chao Xu, Xinyue Wei, Linghao Chen, Chong Zeng, and Hao Su. Zero123++: a single image to consistent multi-view diffusion base model. arXiv preprint arXiv:2310.15110, 2023a. Yichun Shi, Peng Wang, Jianglong Ye, Long Mai, Kejie Li, and Xiao Yang. Mvdream: Multi-view diffusion for 3d generation. In The Tw...

work page internal anchor Pith review arXiv
[13]

Collaborative control for geometry-conditioned pbr image generation

Shimon Vainer, Mark Boss, Mathias Parger, Konstantin Kutsy, Dante De Nigris, Ciara Rowles, Nicolas Perony, and Simon Donn´e. Collaborative control for geometry-conditioned pbr image generation. In Proceedings of European Conference on Computer Vision , pp. 127–145, 2024a. Shimon Vainer, Konstantin Kutsy, Dante De Nigris, Ciara Rowles, Slava Elizarov, and ...

work page arXiv
[14]

arXiv preprint arXiv:2312.02201 , year=

Peng Wang and Yichun Shi. Imagedream: Image-prompt multi-view diffusion for 3d generation. arXiv preprint arXiv:2312.02201,

work page arXiv
[15]

Scaling mesh generation via compressive tokenization

Haohan Weng, Zibo Zhao, Biwen Lei, Xianghui Yang, Jian Liu, Zeqiang Lai, Zhuo Chen, Yuhong Liu, Jie Jiang, Chunchao Guo, et al. Scaling mesh generation via compressive tokenization. arXiv preprint arXiv:2411.07025,

work page arXiv
[16]

Texro: generating delicate textures of 3d models by recursive optimization

Jinbo Wu, Xing Liu, Chenming Wu, Xiaobo Gao, Jialun Liu, Xinqi Liu, Chen Zhao, Haocheng Feng, Errui Ding, and Jingdong Wang. Texro: generating delicate textures of 3d models by recursive optimization. arXiv preprint arXiv:2403.15009, 2024a. Shuang Wu, Youtian Lin, Feihu Zhang, Yifei Zeng, Jingxi Xu, Philip Torr, Xun Cao, and Yao Yao. Direct3d: Scalable im...

work page arXiv 2023
[17]

Structured 3D Latents for Scalable and Versatile 3D Generation

Jianfeng Xiang, Zelong Lv, Sicheng Xu, Yu Deng, Ruicheng Wang, Bowen Zhang, Dong Chen, Xin Tong, and Jiaolong Yang. Structured 3d latents for scalable and versatile 3d generation. arXiv preprint arXiv:2412.01506,

work page internal anchor Pith review arXiv
[18]

Matlaber: Material-aware text-to-3d via latent brdf auto-encoder

Xudong Xu, Zhaoyang Lyu, Xingang Pan, and Bo Dai. Matlaber: Material-aware text-to-3d via latent brdf auto-encoder. arXiv preprint arXiv:2308.09278,

work page arXiv
[19]

Hunyuan3d-1.0: A unified framework for text-to-3d and image-to-3d generation

Xianghui Yang, Huiwen Shi, Bowen Zhang, Fan Yang, Jiacheng Wang, Hongxu Zhao, Xinhai Liu, Xinzhou Wang, Qingxiang Lin, Jiaao Yu, et al. Hunyuan3d-1.0: A unified framework for text-to-3d and image-to-3d generation. arXiv preprint arXiv:2411.02293,

work page arXiv
[20]

Shapegpt: 3d shape generation with a unified multi-modal language model

Fukun Yin, Xin Chen, Chi Zhang, Biao Jiang, Zibo Zhao, Jiayuan Fan, Gang Yu, Taihao Li, and Tao Chen. Shapegpt: 3d shape generation with a unified multi-modal language model. arXiv preprint arXiv:2311.17618,

work page arXiv
[21]

Paint3d: Paint anything 3d with lighting-less texture diffusion models

Xianfang Zeng, Xin Chen, Zhongqi Qi, Wen Liu, Zibo Zhao, Zhibin Wang, Bin Fu, Yong Liu, and Gang Yu. Paint3d: Paint anything 3d with lighting-less texture diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pp. 4252–4262, 2024a. Zheng Zeng, Valentin Deschaintre, Iliyan Georgiev, Yannick Hold-Geoffroy, Y...

work page 2024
[22]

Texpainter: Generative mesh texturing with multi-view consistency

13 Hongkun Zhang, Zherong Pan, Congyi Zhang, Lifeng Zhu, and Xifeng Gao. Texpainter: Generative mesh texturing with multi-view consistency. In ACM SIGGRAPH 2024 Conference Papers , pp. 1–11, 2024a. Longwen Zhang, Ziyu Wang, Qixuan Zhang, Qiwei Qiu, Anqi Pang, Haoran Jiang, Wei Yang, Lan Xu, and Jingyi Yu. Clay: A controllable large-scale generative model ...

work page 2024
[23]

Hunyuan3D 2.0: Scaling Diffusion Models for High Resolution Textured 3D Assets Generation

Zibo Zhao, Zeqiang Lai, Qingxiang Lin, Yunfei Zhao, Haolin Liu, Shuhui Yang, Yifei Feng, Mingxin Yang, Sheng Zhang, Xianghui Yang, et al. Hunyuan3d 2.0: Scaling diffusion models for high resolution textured 3d assets generation. arXiv preprint arXiv:2501.12202,

work page internal anchor Pith review Pith/arXiv arXiv
[24]

Uni3d: Exploring unified 3d representation at scale

Junsheng Zhou, Jinsheng Wang, Baorui Ma, Yu-Shen Liu, Tiejun Huang, and Xinlong Wang. Uni3d: Exploring unified 3d representation at scale. arXiv preprint arXiv:2310.06773,

work page arXiv