Pith. sign in

REVIEW 4 major objections 8 minor 1 cited by

Consistent Zero-shot 3D Texture Synthesis Using Geometry-aware Diffusion and Temporal Video Models

T0 review · 4 major / 8 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read VideoTex treats 3D texture synthesis as a video-to-video generation task over a camera orbit, using geometry conditions and component-wise UV inpainting to keep textures consistent across all viewpoints.

desk verdict A practical texture pipeline with a plausible qualitative win, but the paper never shows that the video prior is what does the work. read the letter →

arxiv 2506.20946 v1 pith:BAG7UPYW submitted 2025-06-26 cs.GR cs.AIcs.CV

Donggoo Kang , Jangyeong Kim , Dasol Jeong , Junyoung Choi , Jeonga Wi , Hyunmin Lee , Joonho Gwon , Joonki Paik This is my paper · ORCID
classification cs.GRcs.AIcs.CV
keywords 3Dtexturesynthesiszero-shottexturingvideodiffusiontemporalconsistencygeometry-awareconditioningUVinpaintingControlNettext-to-texture
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

VideoTex argues that text-driven texturing of a static 3D mesh should be treated not as a set of independent fixed-view generations but as a temporally consistent video-to-video generation problem. The paper renders an untextured mesh along a 360-degree orbit, conditions a diffusion video generator on normal, depth, and edge maps, reverse-projects each frame, and fuses the frames with a squared confidence weight so overlapping texels do not over-smooth. A second stage inpaints occluded regions component-by-component in UV space, treating reusable mesh parts as semantic units. If the claims hold, zero-shot texture synthesis from a prompt can avoid the seams, distortions, and Janus artifacts of fixed-viewpoint methods, and the reported experiments indicate that VideoTex outperforms Text2Tex, Paint3D, and Meshy on fidelity, consistency, alignment, and generation time.

What carries the argument

The load-bearing machinery is the pairing of an orbit video-to-video generation pass with a UV-domain component inpainting pass. The orbit path $O(t) = (r\cos(2\pi t/T), r\sin(2\pi t/T), z)$ creates evenly spaced viewpoints whose rendered normal, depth, and Canny edge maps act as ControlNet conditions (strength 0.7) on an SDXL backbone with an AnimateDiff-style temporal module, so the same texture identity is maintained frame to frame. Reverse projection assigns each frame's colors to texels, and squared confidence blending $w(v)^\alpha$ with $\alpha = 8$ gives sharper seams by trusting only viewpoints facing each surface patch. Finally, component-wise UV diffusion, trained on 37,979 curated Objaverse meshes with Cap3D captions, inpaints the untextured UV regions while preserving the semantic boundaries between reusable components.

What would settle it

Render any textured mesh produced by VideoTex from a dense set of novel viewpoints and measure the color discontinuity across UV chart boundaries and the variance of matched texels between neighboring orbit frames; if those discontinuities are not lower than a fixed-view baseline on a shape class the model was not trained on, the central claim fails.

Watch

Extended reading notes

Core claim

The paper's central claim is that a video diffusion prior can carry the global context that fixed-view texture generators lack, producing a complete UV texture without any per-asset optimization. Concretely, VideoTex defines an orbit path around the mesh, feeds a rendered sequence of geometry-conditioned frames into a video-to-video generator, and maps the temporally coherent result back into UV space using normal-based confidence weights raised to a power alpha. The same pipeline then inpaints the unseen, occluded regions in the UV domain with a component-wise diffusion model that is trained to respect the semantic grouping of mesh parts. The authors report the lowest Kernel Inception Distance (KID) among the compared methods, and a user study with 30 expert evaluators favors VideoTex on texture quality, consistency, and alignment.

Load-bearing premise

Everything rests on the temporal video prior keeping texture identity consistent across widely separated orbit viewpoints; if the video model drifts or fails on a shape it has never seen, the per-view projection and blending cannot repair the inconsistency.

Editorial extensions

If this is right

  • VideoTex offers a practical zero-shot pipeline: texture a fresh mesh from a text prompt in roughly 32 seconds on a single 4090, with no per-object optimization.
  • UV seams and occluded regions inherit temporal consistency from the video prior, so the final asset stays stable when viewed from arbitrary angles.
  • Because each component is inpainted separately in UV space, repeated or semantically meaningful parts keep a coherent appearance even when their UV charts are fragmented.
  • Frame rate becomes a tunable knob: 8 frames per orbit already match higher frame rates, so users can trade speed for quality without hurting KID.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: the squared-confidence blending rule suggests a generic anti-over-smoothing recipe for any multi-view back-projection, including neural-field or Gaussian-splatting renderers, not just UV-mapped meshes.
  • Because the temporal prior is the source of consistency, the method's ceiling tracks video-model quality: a closed-source model with stronger multi-view consistency would likely raise fidelity further, but the open AnimateDiff backbone limits what the current article can show.
  • A testable extension: apply VideoTex to articulated or deformable meshes where the orbit condition is combined with a time-varying pose, and measure whether the video prior maintains texture identity under motion.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 8 minor

Summary. The paper introduces VideoTex, a zero-shot 3D texture synthesis method that frames texture generation as a geometry-conditioned video-to-video problem. It renders an orbit of normal, depth, and edge maps, feeds them along with a text prompt to SDXL augmented with ControlNets and an AnimateDiff motion module, reverse-projects the generated frames onto the UV map using squared-confidence blending (Eqs. 7-8), and finally fills occluded regions with a component-wise UV diffusion model trained on ~38K curated Objaverse assets. The authors report superior KID and user-study scores compared to Text2Tex, Paint3D, and Meshy, and conclude that the video prior yields temporally stable, seam-aware textures.

Significance. If the results hold, VideoTex would be a practical zero-shot baseline for text-driven texturing without per-asset optimization, and the idea of using a temporal video prior for multi-view consistency is timely. The paper ships no code or data and does not provide statistical support for its claims; nevertheless, the proposed pipeline is well-motivated and the component-wise UV inpainting direction is interesting. The main strengths are the explicit use of a video diffusion prior, the orbit-based V2V conditioning, and the large curated training set for UV inpainting. However, the lack of a control isolating the video prior and the weak quantitative evaluation currently prevent the central claims from being accepted.

major comments (4)
  1. [3.3-3.4, Table 1] The paper attributes its consistency and quality improvements to the temporal video prior, but nowhere isolates this mechanism. Please add an experiment that runs the same pipeline with per-frame independent SDXL+ControlNet sampling (no AnimateDiff), holding conditions, seeds, reverse projection, and alpha=8 fixed. Given that Eq. (7) with alpha=8 approximates hard per-texel view selection, the observed seams/stability could be entirely due to the blending rule rather than to temporal coherence. This ablation is load-bearing for the paper's main novelty.
  2. [4.2, Table 1] The quantitative evaluation is statistically inconclusive. KID values are given without error bars, the number of test assets, rendered viewpoints, or seeds; the user-study percentages (54%, 90%, 58%) involve 30 participants and 20 assets but lack confidence intervals or tests of significance. The abstract's claim of outperformance is therefore not supported by the evidence. Please report the standard deviation across multiple GPU seeds (at least 3) and provide a significance test for the user study (e.g., a paired permutation test).
  3. [3.5, 4.3] The contribution of the component-wise UV diffusion is not isolated. The ablations in Section 4.3 only vary framerate and blending alpha; there is no comparison against the same pipeline with a generic UV inpainter (e.g., ControlNet without the component map or an off-the-shelf SDXL inpainting model). Since this is the paper's second major contribution, please report an ablation that removes the component map or replaces the trained ControlNet.
  4. [4.2, Supplementary I] The evaluation is narrow: only three baselines are used, one commercial (Meshy without a reference), and the test set is limited to 20 assets. The related work discusses TexFusion, RoCoTex, Meta 3D TextureGen, and TexturE; if code is unavailable for some, at least implement a per-frame ControlNet baseline (see comment 1) and/or report results for a larger set of assets with per-asset KID breakdown. The current evidence supports a qualitative demo, not the claim of outperformance over the state of the art.
minor comments (8)
  1. [4.2] The sentence 'We experiment with SeamlessTex using various texturing methods' contains a typo; presumably it should say 'VideoTex' or 'seamless texture synthesis.'
  2. [Eq. (6)] The angle theta is not defined; please state that theta is the angle between the surface normal n_v and the viewing direction n_view.
  3. [4.1] The phrase 'We used a frame rate of 8' is ambiguous; the ablation varies the number of frames (4, 8, 16, 24), so use 'video length' or 'frame count'.
  4. [Table 1(b)] The heading 'Effect of frame rate' should be 'Effect of number of frames' to match the values 4, 8, 16, 24.
  5. [References] References [15] (Jarzynski, free energy equality) and [21] (Neal, annealed importance sampling) are unrelated to video generation; these citation errors should be removed or replaced.
  6. [4.2-4.3] Meshy is used as a baseline without a reference; please provide a citation for the tool.
  7. [4.3] In the text, 'Figure 10, 11' is cited for over-smoothing, but it is unclear which visual evidence supports that claim; please clarify the pointer.
  8. [Abstract/4.2] The paper claims 'smoother transitions across UV boundaries' but does not report a seam-specific metric; consider adding a quantitative seam analysis or at least a zoomed UV comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: VideoTex is an empirical pipeline whose results are benchmarked against external baselines, with no load-bearing step reducing to its own inputs.

full rationale

The paper contains no formal derivation whose conclusion is equivalent to its premises. The pipeline is described by generative equations (Eqs. 3, 5, 7, 8, 9) that formulate conditioning, blending, and inpainting operations, but none of these defines the reported evaluation quantities (KID, user-study scores) in terms of the method's own fitted parameters. The squared-confidence blending in Eqs. 7-8 is a post-processing choice, and the tuning of alpha=8 and ControlNet strengths is standard hyperparameter selection, not a fitted-input-called-prediction, because the comparisons are against external methods rather than against quantities constructed from the fitted values. The only self-citation is RoCoTex [17] in Related Work, which is descriptive and not load-bearing for the paper's central claim. The absence of an ablation isolating AnimateDiff from per-frame SDXL sampling is a missing control that weakens the attribution of temporal consistency to the video model, but that is an experimental-design limitation, not circularity by the paper's own definitions or equations. The empirical claims are tested against external baselines (Text2Tex, Paint3D, Meshy) and are therefore self-contained in the sense relevant to circularity analysis.

Assumptions & free parameters 4 free parameters · 6 assumptions · 0 invented entities

The pipeline rests on several empirical transfer assumptions rather than formal derivation. The numbers fitted by hand (alpha, ControlNet strengths, frame count, orbit parameters) are listed as free parameters; no invented entities are introduced.

free parameters (4)
  • alpha (squared confidence blending exponent) = alpha = 8
    Chosen by hand to sharpen per-view confidence maps; Section 4.3 shows ablations over alpha but no automatic selection. It affects final texture sharpness and seam behavior.
  • ControlNet control strengths = 0.7 for depth/normal/edge, 0.5 for UV component
    Set manually in Section 4.1; no sweep or sensitivity analysis is reported.
  • Video frame count for the orbit sequence = 8 frames in the default reported table; tested at 4, 8, 16, and 24
    The orbit resolution is a design choice that trades runtime against viewpoint coverage; higher frame counts increase time without lowering KID.
  • Orbit radius r and camera height z = not reported
    Equation 4 defines the orbit path but the numeric values used in the experiments are not given, so the exact viewpoint sampling cannot be reproduced.
assumptions (6)
  • domain assumption Pre-trained SDXL plus ControlNets on normal, depth, and edge maps can align generated textures to 3D geometry.
    This is the core geometric-alignment premise of Section 3.2; it relies on ControlNet generalization rather than on a new proof.
  • domain assumption AnimateDiff-style video diffusion produces temporally consistent generations that transfer to a camera orbit around an untextured mesh.
    Section 3.3 assumes that video prior consistency across frames maps to viewpoint consistency across an orbit. This is the main transfer assumption and is not formally justified.
  • domain assumption 3D assets are composed of reusable semantic components whose UV layouts carry semantic information useful for inpainting.
    Supports the component-wise UV diffusion strategy in Section 3.5 and the Objaverse curation in Supplementary Section F.
  • domain assumption Per-view confidence weighting with alpha = 8 is sufficient to fuse multi-view projections without a global optimization step.
    Section 3.4 replaces global texture optimization with a local weighted average; the paper validates this only visually and through KID.
  • domain assumption KID on rendered views and the reported user study are valid proxies for 3D texture quality and consistency.
    Section 4.2 uses KID and a 30-participant user study as the evaluation, but no validity analysis or error bars are provided.
  • domain assumption The curated Objaverse training subset is representative of the evaluation assets.
    Supplementary Section F describes filtering for manually authored assets; no held-out split or train/test overlap analysis is reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Consistent Zero-shot 3D Texture Synthesis Using Geometry-aware Diffusion and Temporal Video Models." pith.science (2026). https://pith.science/paper/BAG7UPYW

@misc{pith2026250620946,
  author       = {Pith},
  title        = {Pith review of: Consistent Zero-shot 3D Texture Synthesis Using Geometry-aware Diffusion and Temporal Video Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BAG7UPYW}},
  note         = {Machine review of arXiv:2506.20946}
}
read the original abstract

Current texture synthesis methods, which generate textures from fixed viewpoints, suffer from inconsistencies due to the lack of global context and geometric understanding. Meanwhile, recent advancements in video generation models have demonstrated remarkable success in achieving temporally consistent videos. In this paper, we introduce VideoTex, a novel framework for seamless texture synthesis that leverages video generation models to address both spatial and temporal inconsistencies in 3D textures. Our approach incorporates geometry-aware conditions, enabling precise utilization of 3D mesh structures. Additionally, we propose a structure-wise UV diffusion strategy, which enhances the generation of occluded areas by preserving semantic information, resulting in smoother and more coherent textures. VideoTex not only achieves smoother transitions across UV boundaries but also ensures high-quality, temporally stable textures across video frames. Extensive experiments demonstrate that VideoTex outperforms existing methods in texture fidelity, seam blending, and stability, paving the way for dynamic real-time applications that demand both visual quality and temporal coherence.

Figures

Figures reproduced from arXiv: 2506.20946 by the authors.

Figure 1
Figure 1. This paper presents VideoTex, a seamless texture synthesis approach that utilizes the temporal consistency of a video generation model. We also propose a component-based UV diffusion model to achieve semantic-aware UV diffusion. Abstract Current texture synthesis methods, which generate textures from fixed viewpoints, suffer from inconsistencies due to the lack of global context and geometric understanding. Mean￾whi… view at source ↗
Figure 2
Figure 2. Limitations of existing texture synthesis methods: (a) lack of contextual awareness, (b) misalignment and lack of geometric [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of the UV map for a 3D asset composed of [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: UV map created by the 3D modeler (a) and UV map [PITH_FULL_IMAGE:figures/full_fig_p003_4.png]
Figure 5
Figure 5. Figure 5: Overview of VideoTex. The proposed method generates temporally consistent textures in the coarse stage using a video generation [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Confidence map from the first view (left) and the accu [PITH_FULL_IMAGE:figures/full_fig_p005_6.png]
Figure 7
Figure 7. Figure 7: Visualization of confidence maps for square parameters. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png]
Figure 8
Figure 8. Figure 8: Qualitative comparison results with the state-of-the-art methods. [PITH_FULL_IMAGE:figures/full_fig_p007_8.png]
Figure 9
Figure 9. Figure 9: UV texture map comparison results [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Visualization of accumulated confidence maps from 8 [PITH_FULL_IMAGE:figures/full_fig_p008_10.png]
Figure 11
Figure 11. Figure 11: Texture results by square parameter α. ates don’t always improve quality. This is due to reduced viewpoint variation at higher framerates, which, even with our blending method, leads to pixel overlap and an over￾smoothing effect on texels. Squared Confidence Blending …
Figure 12
Figure 12. Figure 12: UV texture map comparison results 13 [PITH_FULL_IMAGE:figures/full_fig_p013_12.png]
Figure 13
Figure 13. Figure 13: UV texture map and its corresponding UV component map from the training dataset [PITH_FULL_IMAGE:figures/full_fig_p014_13.png]
Figure 14
Figure 14. Figure 14: Additional experimental results 15 [PITH_FULL_IMAGE:figures/full_fig_p015_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Droplet3D: Commonsense Priors from Videos Facilitate 3D Generation

    cs.CV 2025-08 conditional novelty 6.0 of 10

    A video diffusion backbone fine-tuned on 4M densely captioned 360-degree renderings generates spatially consistent multi-view images for 3D assets from image plus detailed text input.

Reference graph

Works this paper leans on

31 extracted references · 18 canonical work pages · cited by 1 Pith paper

  1. [1]

    Meta 3d texturegen: fast and consistent texture generation for 3d objects.arXiv preprint arXiv:2407.02430, 2024

    Raphael Bensadoun, Yanir Kleiman, Idan Azuri, Omri Harosh, Andrea Vedaldi, Natalia Neverova, and Oran Gafni. Meta 3d texturegen: fast and consistent texture generation for 3d objects.arXiv preprint arXiv:2407.02430, 2024. 2

    arXiv 2024

  2. [2]

    Align your latents: High-resolution video synthesis with la- tent diffusion models

    Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dock- horn, Seung Wook Kim, Sanja Fidler, and Karsten Kreis. Align your latents: High-resolution video synthesis with la- tent diffusion models. InProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition, pages 22563–22575, 2023. 2, 3

    work page 2023

  3. [3]

    Texfusion: Synthesizing 3d textures with text-guided image diffusion models

    Tianshi Cao, Karsten Kreis, Sanja Fidler, Nicholas Sharp, and Kangxue Yin. Texfusion: Synthesizing 3d textures with text-guided image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4169–4181, 2023. 2

    work page 2023

  4. [4]

    Text2tex: Text-driven tex- ture synthesis via diffusion models

    Dave Zhenyu Chen, Yawar Siddiqui, Hsin-Ying Lee, Sergey Tulyakov, and Matthias Nießner. Text2tex: Text-driven tex- ture synthesis via diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 18558–18568, 2023. 2, 6

    work page 2023

  5. [5]

    Gentron: Diffusion trans- formers for image and video generation

    Shoufa Chen, Mengmeng Xu, Jiawei Ren, Yuren Cong, Sen He, Yanping Xie, Animesh Sinha, Ping Luo, Tao Xiang, and Juan-Manuel Perez-Rua. Gentron: Diffusion trans- formers for image and video generation. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6441–6451, 2024. 2, 3

    work page 2024

  6. [6]

    Flatten: optical flow- guided attention for consistent text-to-video editing.arXiv preprint arXiv:2310.05922, 2023

    Yuren Cong, Mengmeng Xu, Christian Simon, Shoufa Chen, Jiawei Ren, Yanping Xie, Juan-Manuel Perez-Rua, Bodo Rosenhahn, Tao Xiang, and Sen He. Flatten: optical flow- guided attention for consistent text-to-video editing.arXiv preprint arXiv:2310.05922, 2023. 2, 3

    arXiv

  7. [7]

    Veo 2: Our state-of-the-art video generation model, 2025

    DeepMind. Veo 2: Our state-of-the-art video generation model, 2025. 8

    work page 2025

  8. [8]

    Objaverse: A universe of annotated 3d objects

    Matt Deitke, Dustin Schwenk, Jordi Salvador, Luca Weihs, Oscar Michel, Eli VanderBilt, Ludwig Schmidt, Kiana Ehsani, Aniruddha Kembhavi, and Ali Farhadi. Objaverse: A universe of annotated 3d objects. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13142–13153, 2023. 6

    work page 2023

Show all 31 references
  1. [9]

    Structure and content-guided video synthesis with diffusion models

    Patrick Esser, Johnathan Chiu, Parmida Atighehchian, Jonathan Granskog, and Anastasis Germanidis. Structure and content-guided video synthesis with diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 7346–7356, 2023. 2, 3

    2023

  2. [10]

    Lumina-t2x: Transforming text into any modality, reso- lution, and duration via flow-based large diffusion transform- ers.arXiv preprint arXiv:2405.05945, 2024

    Peng Gao, Le Zhuo, Ziyi Lin, Chris Liu, Junsong Chen, Ruoyi Du, Enze Xie, Xu Luo, Longtian Qiu, Yuhang Zhang, et al. Lumina-t2x: Transforming text into any modality, reso- lution, and duration via flow-based large diffusion transform- ers.arXiv preprint arXiv:2405.05945, 2024

    2024 arXiv

  3. [11]

    Emu video: Factoriz- ing text-to-video generation by explicit image conditioning

    Rohit Girdhar, Mannat Singh, Andrew Brown, Quentin Du- val, Samaneh Azadi, Sai Saketh Rambhatla, Akbar Shah, Xi Yin, Devi Parikh, and Ishan Misra. Emu video: Factoriz- ing text-to-video generation by explicit image conditioning. arXiv preprint arXiv:2311.10709, 2023. 2, 3

    2023 arXiv

  4. [12]

    Animatediff: Animate your personalized text- to-image diffusion models without specific tuning.arXiv preprint arXiv:2307.04725, 2023

    Yuwei Guo, Ceyuan Yang, Anyi Rao, Zhengyang Liang, Yaohui Wang, Yu Qiao, Maneesh Agrawala, Dahua Lin, and Bo Dai. Animatediff: Animate your personalized text- to-image diffusion models without specific tuning.arXiv preprint arXiv:2307.04725, 2023. 3

    2023 arXiv

  5. [13]

    Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models.Advances in neural information processing systems, 33:6840–6851, 2020. 2, 3

    2020

  6. [14]

    Imagen video: High definition video generation with diffusion mod- els.arXiv preprint arXiv:2210.02303, 2022

    Jonathan Ho, William Chan, Chitwan Saharia, Jay Whang, Ruiqi Gao, Alexey Gritsenko, Diederik P Kingma, Ben Poole, Mohammad Norouzi, David J Fleet, et al. Imagen video: High definition video generation with diffusion mod- els.arXiv preprint arXiv:2210.02303, 2022. 3

    arXiv

  7. [15]

    Nonequilibrium equality for free energy differences.Physical Review Letters, 78(14):2690,

    Christopher Jarzynski. Nonequilibrium equality for free energy differences.Physical Review Letters, 78(14):2690,

  8. [16]

    Flexitex: Enhancing tex- ture generation with visual guidance.arXiv preprint arXiv:2409.12431, 2024

    DaDong Jiang, Xianghui Yang, Zibo Zhao, Sheng Zhang, Jiaao Yu, Zeqiang Lai, Shaoxiong Yang, Chunchao Guo, Xiaobo Zhou, and Zhihui Ke. Flexitex: Enhancing tex- ture generation with visual guidance.arXiv preprint arXiv:2409.12431, 2024. 2

    2024 arXiv

  9. [17]

    Rocotex: A robust method for consistent texture synthesis with diffusion models.arXiv preprint arXiv:2409.19989,

    Jangyeong Kim, Donggoo Kang, Junyoung Choi, Jeonga Wi, Junho Gwon, Jiun Bae, Dumim Yoon, and Junghyun Han. Rocotex: A robust method for consistent texture synthesis with diffusion models.arXiv preprint arXiv:2409.19989,

    arXiv

  10. [18]

    Scalable 3d captioning with pretrained models.Ad- vances in Neural Information Processing Systems, 36, 2024

    Tiange Luo, Chris Rockwell, Honglak Lee, and Justin John- son. Scalable 3d captioning with pretrained models.Ad- vances in Neural Information Processing Systems, 36, 2024. 6

    2024

  11. [19]

    Latent-nerf for shape-guided generation of 3d shapes and textures

    Gal Metzer, Elad Richardson, Or Patashnik, Raja Giryes, and Daniel Cohen-Or. Latent-nerf for shape-guided generation of 3d shapes and textures. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12663–12673, 2023. 2

    2023

  12. [20]

    Clip-mesh: Generating textured meshes from text using pretrained image-text models

    Nasir Mohammad Khalid, Tianhao Xie, Eugene Belilovsky, and Tiberiu Popa. Clip-mesh: Generating textured meshes from text using pretrained image-text models. InSIGGRAPH Asia 2022 conference papers, pages 1–8, 2022. 2

    2022

  13. [21]

    Annealed importance sampling.Statistics and computing, 11:125–139, 2001

    Radford M Neal. Annealed importance sampling.Statistics and computing, 11:125–139, 2001. 2, 3

    2001

  14. [22]

    Video generation models as world simulators.Ope- nAI, 2024

    OpenAI. Video generation models as world simulators.Ope- nAI, 2024. 8

    2024

  15. [23]

    Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis.arXiv preprint arXiv:2307.01952, 2023

    Dustin Podell, Zion English, Kyle Lacey, Andreas Blattmann, Tim Dockhorn, Jonas M ¨uller, Joe Penna, and Robin Rombach. Sdxl: Improving latent diffusion mod- els for high-resolution image synthesis.arXiv preprint arXiv:2307.01952, 2023. 4, 6

    2023 arXiv

  16. [24]

    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. InACM SIGGRAPH 2023 Conference Proceedings, pages 1–11, 2023. 2 9

    2023

  17. [25]

    Magicvideo-v2: Multi- stage high-aesthetic video generation.arXiv preprint arXiv:2401.04468, 2024

    Weimin Wang, Jiawei Liu, Zhijie Lin, Jiangqiao Yan, Shuo Chen, Chetwin Low, Tuyen Hoang, Jie Wu, Jun Hao Liew, Hanshu Yan, et al. Magicvideo-v2: Multi- stage high-aesthetic video generation.arXiv preprint arXiv:2401.04468, 2024. 2, 3

    2024 arXiv

  18. [26]

    Texro: Generating delicate textures of 3d models by recursive optimization.arXiv preprint arXiv:2403.15009, 2024

    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, 2024. 2

    2024 arXiv

  19. [27]

    Tune-a-video: One-shot tuning of image diffusion models for text-to-video generation

    Jay Zhangjie Wu, Yixiao Ge, Xintao Wang, Stan Weixian Lei, Yuchao Gu, Yufei Shi, Wynne Hsu, Ying Shan, Xiaohu Qie, and Mike Zheng Shou. Tune-a-video: One-shot tuning of image diffusion models for text-to-video generation. In Proceedings of the IEEE/CVF International Conference...

    2023

  20. [28]

    Texture generation on 3d meshes with point- uv diffusion

    Xin Yu, Peng Dai, Wenbo Li, Lan Ma, Zhengzhe Liu, and Xiaojuan Qi. Texture generation on 3d meshes with point- uv diffusion. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4206–4216, 2023. 4

    2023

  21. [29]

    Paint3d: Paint anything 3d with lighting-less texture diffusion models.arXiv preprint arXiv:2312.13913,

    Xianfang Zeng. Paint3d: Paint anything 3d with lighting-less texture diffusion models.arXiv preprint arXiv:2312.13913,

    arXiv

  22. [30]

    Show-1: Marrying pixel and latent diffusion models for text-to-video generation.International Journal of Com- puter Vision, pages 1–15, 2024

    David Junhao Zhang, Jay Zhangjie Wu, Jia-Wei Liu, Rui Zhao, Lingmin Ran, Yuchao Gu, Difei Gao, and Mike Zheng Shou. Show-1: Marrying pixel and latent diffusion models for text-to-video generation.International Journal of Com- puter Vision, pages 1–15, 2024. 2, 3

    2024

  23. [31]

    snow cottage

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 3836–3847, 2023. 4, 6 10 Consistent Zero-shot 3D Texture Synthesis Using Geometry-aware ...

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.