Pith. sign in

REVIEW 4 major objections 5 minor 77 references

OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation

T0 review · 4 major / 5 minor · reviewed 2026-08-09 · deepseek-v4-flash

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

Pith's one-line read OmniPhysGS turns each 3D Gaussian into a learnable constitutive particle, so one diffusion-guided pipeline can synthesize physically plausible dynamics for rubber, sand, honey, water, and mixed-material scenes from a text prompt.

desk verdict Solid, incremental 4D-generation paper that learns to pick among 12 constitutive models per neighborhood, but the evaluation never actually measures physical plausibility, so the central claim outruns the evidence. read the letter →

arxiv 2501.18982 v1 pith:CCLRBZET submitted 2025-01-31 cs.CV

Yuchen Lin , Chenguo Lin , Jianjin Xu , Yadong Mu This is my paper · ORCID
classification cs.CV
keywords 3DGaussianSplattingMaterialPointMethodConstitutiveGaussiansmodelsScoreDistillationSamplingvideodiffusionpriorphysics-baseddynamicscenesynthesis4Dgeneration
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

OmniPhysGS claims that a single pipeline can synthesize physically plausible 3D dynamics for heterogeneous materials—rubber, sand, honey, water, and their interactions—without anyone manually setting material parameters. The method upgrades each 3D Gaussian splat into a Constitutive Gaussian, a particle whose local neighborhood chooses one of 12 expert-designed constitutive models according to MLP-predicted logits, and it learns that choice end-to-end from the score-distillation gradient of a pretrained text-to-video diffusion model. The paper argues this is the first learnable constitutive-model formulation for physics-based 3D dynamic synthesis, and it reports roughly 3-16% gains in text-alignment and visual-quality metrics over systems that keep a single preset material model. If the claim holds, text prompts alone could animate scenes involving elastic, viscoelastic, plastic, and fluid materials for games, virtual reality, robotic simulation, and design preview.

What carries the argument

The central object is the Constitutive Gaussian: a 3D Gaussian kernel augmented with learnable constitutive-model parameters $\theta = \{\theta_{el}, \theta_{pl}, \theta_{phy}\}$. Rather than letting a free neural network predict stress and deformation directly, a physics-aware decoder computes material logits for each neighborhood from the encoded kernel features and applies a straight-through hardmax to commit to one expert constitutive model per neighborhood (Eq. 6). The selected expert laws then feed a differentiable Material Point Method, which updates positions and deformation gradients; a vectorized GPU reimplementation makes this long Markov chain trainable, while a grouped, mini-batch training strategy keeps the roughly $10^3$-step simulation stable under SDS guidance.

What would settle it

Record a real object with known material parameters, such as a silicone cylinder with a measured stress-strain curve, reconstruct it as 3D Gaussians, run OmniPhysGS with a contradictory prompt like 'collapses like dry sand,' and compare the simulated deformation with a high-speed video and a finite-element reference; the central claim fails if the output follows the prompt while disagreeing with the measured mechanics.

Watch

Extended reading notes

Core claim

In the Material Point Method, the identity of a material is fully encoded by the hyperelastic energy density $\Psi$, the plasticity return map $\psi$, and the physical parameters $\gamma$. OmniPhysGS makes each of these learnable per Gaussian particle inside a physics-aware decoder, so a neighborhood is assigned one expert model from 12 combinations of three energy functions (fixed corotated, Neo-Hookean, StVK) and four return maps (identity, Drucker-Prager, von Mises, fluid) through a straight-through hardmax operation (Eq. 6). Starting from static Gaussian kernels, a differentiable MPM simulation advances the scene, the simulated states are rendered into a video clip, and the Score Distillation Sampling loss against a pretrained text-to-video diffusion model (Eq. 7) pushes the constitutive-model selection toward the behavior described by the user prompt. The result, as the paper reports, is automatic generation of elastic, viscoelastic, plastic, and fluid dynamics, including interactions between different materials in one scene, without manual material tuning.

Load-bearing premise

The method's only teacher is a pretrained AI video model; if that model produces pretty but physically wrong motion, the learned materials will copy the appearance rather than real physics.

Editorial extensions

If this is right

  • Changing only the text prompt can re-animate the same static scene as rubber, sand, jelly, water, or a mixture, so one template serves many material assignments.
  • Users no longer need to supply Young's modulus, Poisson's ratio, or damping values; the diffusion prior discovers them through SDS gradients.
  • Mixed-material scenes, such as an object falling into water or a rigid collision denting a metal can, stay inside the same MPM formulation instead of requiring per-object hand-tuning.
  • The grouped multi-batch training strategy makes end-to-end optimization of long simulations feasible on a single GPU, a prerequisite for scaling the approach to more complex scenes.
  • The reported solver comparison indicates roughly 75% lower training memory than the previous MPM solver used in the comparisons.

Reading between the lines

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

  • If the learned 12-law palette is the right inductive bias, a continuous family of constitutive laws interpolating among those experts should cover materials outside the palette with the same pipeline.
  • The same setup can double as a probe for whether video diffusion priors encode mechanics: contradictory prompts like 'a steel balloon' would reveal whether the optimizer follows surface appearance or physical knowledge.
  • The per-scene SDS cost is high, so the framework points toward amortized training across many scenes to make material inference from a prompt nearly instant.
  • Because material labels are assigned per neighborhood, the method implicitly segments a scene by material, which could be reused for physics-aware editing or robot manipulation planning.
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 / 5 minor

Summary. The paper proposes OmniPhysGS, a method for generating physics-based 3D dynamics from a static 3D Gaussian Splatting scene and a text prompt. Each Gaussian particle is equipped with a learnable constitutive model, implemented as a per-neighborhood hardmax selection over 12 expert-designed hyperelastic/plasticity constitutive models, and the selection is optimized end-to-end through a differentiable MPM simulator using Score Distillation Sampling from the ModelScope video diffusion model. The authors report quantitative comparisons against PhysDreamer, Physics3D, and DreamPhysics, claiming roughly 3% to 16% improvements in text-alignment and related metrics, together with qualitative results for elastic, viscoelastic, plastic, and fluid materials and multi-material interactions.

Significance. The core idea is timely and valuable: replacing a fixed constitutive model with a learnable, physics-constrained mixture of expert constitutive models inside an MPM-based Gaussian dynamics pipeline is a natural and potentially important extension of existing work. The paper also contributes a vectorized PyTorch MPM solver that substantially reduces training memory (Table 1), includes ablations for the network architecture and training strategy, and plans to release code. If the physical-fidelity claims are supported by stronger evaluation, this would be a useful advance for text-driven 4D content generation. However, the current evidence does not yet establish the central claim of physically plausible material estimation, because the reported metrics do not measure physical correctness.

major comments (4)
  1. [Section 4.1, Tables 2–3, Appendix B.3] The evaluation metrics do not support the paper's central claim of “physically plausible” dynamics. CLIPSIM (Eq. 31) measures text-frame alignment, while DiffSSIM and DiffCLIP (Eq. 32) measure dissimilarity from a randomly initialized model; none of these detects violations of mechanics, and high values can be achieved by visually salient but physically wrong rollouts. Moreover, Tables 2 and 3 report single numbers without error bars or significance tests, so the claimed 3–16% improvements over baselines are not established as reproducible. I ask for physical validation: for example, material-identification accuracy on scenes with known constitutive parameters, comparison of trajectories against a reference MPM rollout, checks of energy/momentum behavior, or a perceptual study focused on physical plausibility, plus variance over random seeds.
  2. [Section 3.2.3, Eq. 7; Section 4.1] The only learning signal for the per-neighborhood material selection is SDS against ModelScope (Eq. 7), and the main evaluation uses CLIPSIM, which is the same kind of text-video alignment. This creates a circularity risk: improved CLIPSIM/Diff* scores may reflect better alignment with the diffusion prior's appearance rather than correct material behavior. Since the authors already generate scenes with BlenderNerf (Section 4.1), they have ground-truth material labels available; I ask for a direct test in which the learned expert assignments and continuous physical parameters are compared with the known materials, and for an evaluation on held-out prompts or tasks that is not based on the same diffusion model used for training.
  3. [Section 3.2.2, Eq. 6] The hardmax assignment fixes one expert constitutive model per FPS/kNN neighborhood computed from initial positions. For scenes with fluids, large deformations, or interpenetrating objects, particles from different materials can move into the same neighborhood, so this fixed homogeneous partition may misassign material identity during the simulation. The paper does not ablate or verify this assumption. A concrete test would be to report per-particle material labels over time in a multi-material scene and to compare a variant that recomputes neighborhoods periodically.
  4. [Section 3.2.1–3.2.2, Eq. 4–6] The formulation introduces learnable continuous physical parameters gamma (Eq. 4), yet the architecture description in Section 3.2.2 only accounts for discrete expert selection via logits; no details are given for how gamma is parameterized, initialized, or optimized, and no experiment reports the learned values. This makes the “automatic material estimation” claim under-specified and hard to reproduce, and it leaves open whether the discrete expert selection alone, rather than continuous parameter estimation, drives the reported gains.
minor comments (5)
  1. [Eq. 7] The chain-rule notation ∂V/∂x,F and ∂x,F/∂θ is ambiguous; please spell out the products of Jacobians, such as (∂V/∂x)(∂x/∂θ) + (∂V/∂F)(∂F/∂θ).
  2. [Figure 3] The caption contains a typo: “Consititutive” should be “Constitutive”.
  3. [Section 4.1, Appendix B.3] Please specify how the randomly initialized model for DiffSSIM and DiffCLIP is generated (same scene, same prompt, number of seeds), since the random baseline determines both metrics.
  4. [Appendix B.4, Listing 1] In the training-strategy code snippet, clarify whether the diffusion loss is computed over the whole stage clip or over the mini-batches described in Eq. 8; the current indentation and variable reuse make this unclear.
  5. [Section 4.2] The claim that all baselines “tend to predict homogeneous material properties” is supported only by qualitative figures; a quantitative measure of per-object material parameter estimates would strengthen the comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the constitutive-model selection is genuinely learned from video-diffusion SDS, and the weak evaluation metrics are a validation gap rather than a circular reduction.

full rationale

The paper's derivation chain is self-contained rather than circular. Constitutive Gaussians (Section 3.2.1) extend each Gaussian with learnable hyperelastic, plastic-return, and physical-parameter networks; Eq. 6 selects one of 12 expert constitutive models per neighborhood by hardmax over MLP logits; Eq. 7 back-propagates the SDS loss from the pretrained video diffusion model through the MPM simulation. No equation defines its target in terms of its conclusion, and no result is imported from the authors' own prior work: the 12 constitutive models are taken from external references (Ma et al., 2023; Zong et al., 2023; Xie et al., 2024) and the comparisons are against reimplemented baselines. The evaluation in Section 4.1 uses CLIPSIM, DiffSSIM, and DiffCLIP, which measure text-video alignment and deviation from a random initialization; these metrics do not establish physical correctness, so the paper's 'physically plausible' claim is under-supported, but that is a benchmarking and validity limitation rather than a circularity, because no fitted parameter is renamed as an independent prediction and no evaluation quantity is identical by construction to the training objective. Hence no specific circular step can be quoted with the required reduction.

Assumptions & free parameters 1 free parameters · 5 assumptions · 1 invented entities

The central claim relies on the fidelity of MPM, the coverage of the 12 hand-selected constitutive models, the physical validity of the diffusion-model supervision, and the homogeneous-neighborhood assumption. The material category logits are the only significant free parameters; they are fitted via SDS without independent physical labels.

free parameters (1)
  • per-neighborhood material category logits = selected expert constitutive model per neighborhood, 12-way hardmax, not reported numerically
    Optimized by SDS to match a pretrained video diffusion model; no ground-truth material labels or physical measurements are used to fit them. Central to the claim of automatic material estimation.
assumptions (5)
  • domain assumption The Material Point Method with MLS-MPM accurately approximates the continuum-mechanics conservation equations for the materials in the scenes.
    Section A.2 uses MLS-MPM as the simulator; if the discretization or boundary conditions are inaccurate for fluids or plasticity, generated dynamics are not physically correct.
  • domain assumption The 12 expert-designed constitutive models (3 elastic energy functions multiplied by 4 plasticity return functions) cover the range of materials the model is claimed to handle.
    Section 3.2.2 and Appendix A.4; the set is fixed and drawn from prior work. The paper itself notes in Limitations that the menu is limited to several representative materials.
  • domain assumption The pretrained text-to-video diffusion model (ModelScope) provides physically reliable supervision for material behavior.
    Section 3.2.3: the material selections are learned solely by minimizing the SDS loss against this model; no physics ground truth is used. If the diffusion prior is not physically accurate, the learned material weights are wrong.
  • domain assumption Each FPS/kNN neighborhood contains homogeneous material, and the partition computed at the initial state remains valid during deformation.
    Section 3.2.2 states "Assuming homogeneous material properties within each neighborhood"; the partition is fixed before MPM simulation, so material boundaries may be misclassified after large deformation.
  • standard math The straight-through estimator gives usable gradients through the hardmax material selection.
    Section 3.2.2 uses straight-through estimation, a known technique from Bengio et al.; it is an approximation but does not by itself challenge the central claim.
invented entities (1)
  • Constitutive Gaussian
    purpose: Extends a standard 3D Gaussian kernel with a learnable constitutive model (hyperelastic energy, plasticity return, and physical parameters) so that MPM can assign different material behavior per particle and neighborhood.
    It is an internal representation introduced by the paper; no independent falsifiable handle outside the paper. Its value is measured only through generated video metrics, not through physical measurements.

how reviews work

0 comments
Cite this review

Pith. "Pith review of OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation." pith.science (2026). https://pith.science/paper/CCLRBZET

@misc{pith2026250118982,
  author       = {Pith},
  title        = {Pith review of: OmniPhysGS: 3D Constitutive Gaussians for General Physics-Based Dynamics Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CCLRBZET}},
  note         = {Machine review of arXiv:2501.18982}
}
read the original abstract

Recently, significant advancements have been made in the reconstruction and generation of 3D assets, including static cases and those with physical interactions. To recover the physical properties of 3D assets, existing methods typically assume that all materials belong to a specific predefined category (e.g., elasticity). However, such assumptions ignore the complex composition of multiple heterogeneous objects in real scenarios and tend to render less physically plausible animation given a wider range of objects. We propose OmniPhysGS for synthesizing a physics-based 3D dynamic scene composed of more general objects. A key design of OmniPhysGS is treating each 3D asset as a collection of constitutive 3D Gaussians. For each Gaussian, its physical material is represented by an ensemble of 12 physical domain-expert sub-models (rubber, metal, honey, water, etc.), which greatly enhances the flexibility of the proposed model. In the implementation, we define a scene by user-specified prompts and supervise the estimation of material weighting factors via a pretrained video diffusion model. Comprehensive experiments demonstrate that OmniPhysGS achieves more general and realistic physical dynamics across a broader spectrum of materials, including elastic, viscoelastic, plastic, and fluid substances, as well as interactions between different materials. Our method surpasses existing methods by approximately 3% to 16% in metrics of visual quality and text alignment.

Figures

Figures reproduced from arXiv: 2501.18982 by the authors.

Figure 1
Figure 1. Comparison with Previous Methods. Existing methods rely on handcrafted or narrowly restrictive physical models (e.g., pure elasticity) that limit generalizability. Our method introduces Constitutive Gaussians to better represent physical materials, thus achieving more automatic and physically plausible dynamic synthesis of various materials within a unified framework. Our task is to synthesize general physics-based … view at source ↗
Figure 2
Figure 2. Method Overview. OMNIPHYSGS extends 3D Gaussians with learnable constitutive models, introducing Constitutive Gaussians to the differentiable Material Point Method (MPM). A pre-trained video diffusion model is used to guide the training with Score Distillation Sampling. During the training process, we propose Constitutive Gaussians that are trained to predict the optimal material properties of each Gaussian particle… view at source ↗
Figure 3
Figure 3. Constitutive Gaussian Network. The network architecture of Constitutive Gaussians consists of a 3D feature encoder and a physical-aware decoder. Expert-designed constitutive models are integrated into the decoder to guide the learning process, effectively avoiding the convergence issues faced by vanilla neural networks. where L is the loss function to be introduced in Section 3.2.3. Different from previous works (Zh… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Qualitative visualizations of 3D dynamic synthesis for a single object in different materi [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: Qualitative visualizations of 3D dynamic synthesis for multiple objects in different mate [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: Visualization results of motion generalization. [PITH_FULL_IMAGE:figures/full_fig_p021_6.png]
Figure 7
Figure 7. Figure 7: Qualitative visualizations of 3D dynamic synthesis for a single object with our method. [PITH_FULL_IMAGE:figures/full_fig_p022_7.png]
Figure 8
Figure 8. Figure 8: Qualitative visualizations of 3D dynamic synthesis for multiple objects with our method. [PITH_FULL_IMAGE:figures/full_fig_p023_8.png]
Figure 9
Figure 9. Figure 9: Qualitative visualizations of 3D dynamic synthesis for a single object in different materi [PITH_FULL_IMAGE:figures/full_fig_p024_9.png]
Figure 10
Figure 10. Figure 10: Qualitative visualizations of 3D dynamic synthesis for a single object in different mate [PITH_FULL_IMAGE:figures/full_fig_p025_10.png]
Figure 11
Figure 11. Figure 11: Qualitative visualizations of 3D dynamic synthesis for a single object in different mate [PITH_FULL_IMAGE:figures/full_fig_p026_11.png]
Figure 12
Figure 12. Figure 12: Qualitative visualizations of 3D dynamic synthesis for multiple objects in different ma [PITH_FULL_IMAGE:figures/full_fig_p027_12.png]
Figure 13
Figure 13. Figure 13: Qualitative visualizations of 3D dynamic synthesis for multiple objects in different ma [PITH_FULL_IMAGE:figures/full_fig_p028_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

77 extracted references · 51 canonical work pages

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    4d-fy: Text-to-4d generation using hybrid score distillation sampling

    Sherwin Bahmani, Ivan Skorokhodov, Victor Rong, Gordon Wetzstein, Leonidas Guibas, Peter Wonka, Sergey Tulyakov, Jeong Joon Park, Andrea Tagliasacchi, and David B Lindell. 4d-fy: Text-to-4d generation using hybrid score distillation sampling. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    work page 2024

  3. [3]

    Real-time subspace integration for st

    Jernej Barbi c and Doug L James. Real-time subspace integration for st. venant-kirchhoff deformable models. ACM transactions on graphics (TOG), 2005

    work page 2005

  4. [5]

    ipoke: Poking a still image for controlled stochastic video synthesis

    Andreas Blattmann, Timo Milbich, Michael Dorkenwald, and Bj \"o rn Ommer. ipoke: Poking a still image for controlled stochastic video synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), 2021

    work page 2021

  5. [6]

    Nonlinear continuum mechanics for finite element analysis

    Javier Bonet and Richard D Wood. Nonlinear continuum mechanics for finite element analysis. Cambridge university press, 1997

    work page 1997

  6. [7]

    Gasp: Gaussian splatting for physic-based simulations

    Piotr Borycki, Weronika Smolak, Joanna Waczy \'n ska, Marcin Mazur, S awomir Tadeja, and Przemys aw Spurek. Gasp: Gaussian splatting for physic-based simulations. arXiv preprint arXiv:2409.05819, 2024

    arXiv 2024

  7. [8]

    Hybrid discrete-continuum modeling of shear localization in granular media

    Peter Yichen Chen, Maytee Chantharayukhonthorn, Yonghao Yue, Eitan Grinspun, and Ken Kamrin. Hybrid discrete-continuum modeling of shear localization in granular media. Journal of the Mechanics and Physics of Solids, 2021

    work page 2021

  8. [9]

    Motion-conditioned diffusion model for controllable video synthesis

    Tsai-Shien Chen, Chieh Hubert Lin, Hung-Yu Tseng, Tsung-Yi Lin, and Ming-Hsuan Yang. Motion-conditioned diffusion model for controllable video synthesis. arXiv preprint arXiv:2304.14404, 2023 a

    arXiv 2023

Show all 77 references
  1. [10]

    Control-a-video: Controllable text-to-video generation with diffusion models

    Weifeng Chen, Yatai Ji, Jie Wu, Hefeng Wu, Pan Xie, Jiashi Li, Xin Xia, Xuefeng Xiao, and Liang Lin. Control-a-video: Controllable text-to-video generation with diffusion models. arXiv preprint arXiv:2305.13840, 2023 b

    2023 arXiv

  2. [11]

    Livephoto: Real image animation with text-guided motion control

    Xi Chen, Zhiheng Liu, Mengting Chen, Yutong Feng, Yu Liu, Yujun Shen, and Hengshuang Zhao. Livephoto: Real image animation with text-guided motion control. arXiv preprint arXiv:2312.02928, 2023 c

    2023 arXiv

  3. [12]

    Soil mechanics and plastic analysis or limit design

    Daniel Charles Drucker and William Prager. Soil mechanics and plastic analysis or limit design. Quarterly of applied mathematics, 1952

    1952

  4. [13]

    Gaussian splashing: Unified particles for versatile motion synthesis and rendering, 2024

    Yutao Feng, Xiang Feng, Yintong Shang, Ying Jiang, Chang Yu, Zeshun Zong, Tianjia Shao, Hongzhi Wu, Kun Zhou, Chenfanfu Jiang, and Yin Yang. Gaussian splashing: Unified particles for versatile motion synthesis and rendering, 2024

    2024

  5. [14]

    Sync4d: Video guided controllable dynamics for physics-based 4d generation

    Zhoujie Fu, Jiacheng Wei, Wenhao Shen, Chaoyue Song, Xiaofeng Yang, Fayao Liu, Xulei Yang, and Guosheng Lin. Sync4d: Video guided controllable dynamics for physics-based 4d generation. arXiv preprint arXiv:2405.16849, 2024

    2024 arXiv

  6. [15]

    Gpu optimization of material point methods

    Ming Gao, Xinlei Wang, Kui Wu, Andre Pradhana, Eftychios Sifakis, Cem Yuksel, and Chenfanfu Jiang. Gpu optimization of material point methods. ACM Transactions on Graphics (TOG), 2018

    2018

  7. [16]

    A first course in continuum mechanics

    Oscar Gonzalez and Andrew M Stuart. A first course in continuum mechanics. Cambridge University Press, 2008

    2008

  8. [17]

    Controllable video generation with sparse trajectories

    Zekun Hao, Xun Huang, and Serge Belongie. Controllable video generation with sparse trajectories. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018

    2018

  9. [18]

    Video diffusion models

    Jonathan Ho, Tim Salimans, Alexey Gritsenko, William Chan, Mohammad Norouzi, and David J Fleet. Video diffusion models. Advances in Neural Information Processing Systems (NeurIPS), 2022

    2022

  10. [19]

    A moving least squares material point method with displacement discontinuity and two-way rigid body coupling

    Yuanming Hu, Yu Fang, Ziheng Ge, Ziyin Qu, Yixin Zhu, Andre Pradhana, and Chenfanfu Jiang. A moving least squares material point method with displacement discontinuity and two-way rigid body coupling. ACM Transactions on Graphics (TOG), 2018

    2018

  11. [20]

    Dreamphysics: Learning physical properties of dynamic 3d gaussians with video diffusion priors

    Tianyu Huang, Yihan Zeng, Hui Li, Wangmeng Zuo, and Rynson WH Lau. Dreamphysics: Learning physical properties of dynamic 3d gaussians with video diffusion priors. arXiv preprint arXiv:2406.01476, 2024

    2024 arXiv

  12. [21]

    Plasticinelab: A soft-body manipulation benchmark with differentiable physics

    Zhiao Huang, Yuanming Hu, Tao Du, Siyuan Zhou, Hao Su, Joshua B Tenenbaum, and Chuang Gan. Plasticinelab: A soft-body manipulation benchmark with differentiable physics. arXiv preprint arXiv:2104.03311, 2021

    2021 arXiv

  13. [22]

    The affine particle-in-cell method

    Chenfanfu Jiang, Craig Schroeder, Andrew Selle, Joseph Teran, and Alexey Stomakhin. The affine particle-in-cell method. ACM Transactions on Graphics (TOG), 2015

    2015

  14. [23]

    The material point method for simulating continuum materials, 2016

    Chenfanfu Jiang, Craig Schroeder, Joseph Teran, Alexey Stomakhin, and Andrew Selle. The material point method for simulating continuum materials, 2016

    2016

  15. [24]

    Animate3d: Animating any 3d model with multi-view video diffusion

    Yanqin Jiang, Chaohui Yu, Chenjie Cao, Fan Wang, Weiming Hu, and Jin Gao. Animate3d: Animating any 3d model with multi-view video diffusion. In Advances in neural information processing systems (NeurIPS), 2024 a

    2024

  16. [25]

    Consistent4d: Consistent 360 dynamic object generation from monocular video

    Yanqin Jiang, Li Zhang, Jin Gao, Weiming Hu, and Yao Yao. Consistent4d: Consistent 360 dynamic object generation from monocular video. In International Conference on Learning Representations (ICLR), 2024 b

    2024

  17. [26]

    Dreampose: Fashion image-to-video synthesis via stable diffusion

    Johanna Karras, Aleksander Holynski, Ting-Chun Wang, and Ira Kemelmacher-Shlizerman. Dreampose: Fashion image-to-video synthesis via stable diffusion. In IEEE/CVF International Conference on Computer Vision (ICCV), 2023

    2023

  18. [27]

    3d gaussian splatting for real-time radiance field rendering

    Bernhard Kerbl, Georgios Kopanas, Thomas Leimk \"u hler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics (TOG), 2023

    2023

  19. [28]

    Adam: A method for stochastic optimization

    Diederik P Kingma. Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  20. [29]

    Drucker-prager elastoplasticity for sand animation

    Gergely Kl \'a r, Theodore Gast, Andre Pradhana, Chuyuan Fu, Craig Schroeder, Chenfanfu Jiang, and Joseph Teran. Drucker-prager elastoplasticity for sand animation. ACM Transactions on Graphics (TOG), 2016

    2016

  21. [30]

    Pac-nerf: Physics augmented continuum neural radiance fields for geometry-agnostic system identification

    Xuan Li, Yi-Ling Qiao, Peter Yichen Chen, Krishna Murthy Jatavallabhula, Ming Lin, Chenfanfu Jiang, and Chuang Gan. Pac-nerf: Physics augmented continuum neural radiance fields for geometry-agnostic system identification. arXiv preprint arXiv:2303.05512, 2023

    2023 arXiv

  22. [31]

    Generative image dynamics

    Zhengqi Li, Richard Tucker, Noah Snavely, and Aleksander Holynski. Generative image dynamics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    2024

  23. [32]

    Magic3d: High-resolution text-to-3d content creation

    Chen-Hsuan Lin, Jun Gao, Luming Tang, Towaki Takikawa, Xiaohui Zeng, Xun Huang, Karsten Kreis, Sanja Fidler, Ming-Yu Liu, and Tsung-Yi Lin. Magic3d: High-resolution text-to-3d content creation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio...

    2023

  24. [33]

    Phy124: Fast physics-driven 4d content generation from a single image

    Jiajing Lin, Zhenzhong Wang, Yongjie Hou, Yuzhou Tang, and Min Jiang. Phy124: Fast physics-driven 4d content generation from a single image. arXiv preprint arXiv:2409.07179, 2024

    2024 arXiv

  25. [34]

    Align your gaussians: Text-to-4d with dynamic 3d gaussians and composed diffusion models

    Huan Ling, Seung Wook Kim, Antonio Torralba, Sanja Fidler, and Karsten Kreis. Align your gaussians: Text-to-4d with dynamic 3d gaussians and composed diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    2024

  26. [35]

    Physics3d: Learning physical properties of 3d gaussians via video diffusion

    Fangfu Liu, Hanyang Wang, Shunyu Yao, Shengjun Zhang, Jie Zhou, and Yueqi Duan. Physics3d: Learning physical properties of 3d gaussians via video diffusion. arXiv preprint arXiv:2406.04338, 2024

    2024 arXiv

  27. [36]

    Learning neural constitutive laws from motion observations for generalizable pde dynamics

    Pingchuan Ma, Peter Yichen Chen, Bolei Deng, Joshua B Tenenbaum, Tao Du, Chuang Gan, and Wojciech Matusik. Learning neural constitutive laws from motion observations for generalizable pde dynamics. In International Conference on Machine Learning (ICML), 2023

    2023

  28. [37]

    Warp: A high-performance python framework for gpu simulation and graphics

    Miles Macklin. Warp: A high-performance python framework for gpu simulation and graphics. https://github.com/nvidia/warp, 2022. NVIDIA GPU Technology Conference (GTC)

    2022

  29. [38]

    Srinivasan, Matthew Tancik, Jonathan T

    Ben Mildenhall, Pratul P. Srinivasan, Matthew Tancik, Jonathan T. Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. In European Conference on Computer Vision (ECCV), 2020

    2020

  30. [39]

    o rper im plastisch-deformablen zustand. Nachrichten von der Gesellschaft der Wissenschaften zu G \

    R v Mises. Mechanik der festen k \"o rper im plastisch-deformablen zustand. Nachrichten von der Gesellschaft der Wissenschaften zu G \"o ttingen, Mathematisch-Physikalische Klasse , 1913

    1913

  31. [40]

    Fast dynamic 3d object generation from a single-view video

    Zijie Pan, Zeyu Yang, Xiatian Zhu, and Li Zhang. Fast dynamic 3d object generation from a single-view video. arXiv preprint arXiv:2401.08742, 2024

    2024

  32. [41]

    Dreamfusion: Text-to-3d using 2d diffusion

    Ben Poole, Ajay Jain, Jonathan T Barron, and Ben Mildenhall. Dreamfusion: Text-to-3d using 2d diffusion. In The Eleventh International Conference on Learning Representations (ICLR), 2023

    2023

  33. [42]

    Pointnet: Deep learning on point sets for 3d classification and segmentation

    Charles R Qi, Hao Su, Kaichun Mo, and Leonidas J Guibas. Pointnet: Deep learning on point sets for 3d classification and segmentation. In Proceedings of the IEEE conference on computer vision and pattern recognition (CVPR), 2017 a

    2017

  34. [43]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. Advances in neural information processing systems (NeurIPS), 2017 b

    2017

  35. [44]

    Language-driven physics-based scene synthesis and editing via feature splatting

    Ri-Zhao Qiu, Ge Yang, Weijia Zeng, and Xiaolong Wang. Language-driven physics-based scene synthesis and editing via feature splatting. In European Conference on Computer Vision (ECCV), 2024

    2024

  36. [45]

    BlenderNeRF , 2024

    Maxime Raafat. BlenderNeRF , 2024. URL https://github.com/maximeraafat/BlenderNeRF

    2024

  37. [46]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learning transferable visual models from natural language supervision. In International conference on machine learning (IC...

    2021

  38. [47]

    Dreamgaussian4d: Generative 4d gaussian splatting

    Jiawei Ren, Liang Pan, Jiaxiang Tang, Chi Zhang, Ang Cao, Gang Zeng, and Ziwei Liu. Dreamgaussian4d: Generative 4d gaussian splatting. arXiv preprint arXiv:2312.17142, 2023

    2023 arXiv

  39. [48]

    L4gm: Large 4d gaussian reconstruction model

    Jiawei Ren, Kevin Xie, Ashkan Mirzaei, Hanxue Liang, Xiaohui Zeng, Karsten Kreis, Ziwei Liu, Antonio Torralba, Sanja Fidler, Seung Wook Kim, et al. L4gm: Large 4d gaussian reconstruction model. In Advances in neural information processing systems (NeurIPS), 2024

    2024

  40. [49]

    Animating arbitrary objects via deep motion transfer

    Aliaksandr Siarohin, St \'e phane Lathuili \`e re, Sergey Tulyakov, Elisa Ricci, and Nicu Sebe. Animating arbitrary objects via deep motion transfer. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019 a

    2019

  41. [50]

    First order motion model for image animation

    Aliaksandr Siarohin, St \'e phane Lathuili \`e re, Sergey Tulyakov, Elisa Ricci, and Nicu Sebe. First order motion model for image animation. In Advances in neural information processing systems (NeurIPS), 2019 b

    2019

  42. [51]

    Motion representations for articulated animation

    Aliaksandr Siarohin, Oliver J Woodford, Jian Ren, Menglei Chai, and Sergey Tulyakov. Motion representations for articulated animation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021

    2021

  43. [52]

    Text-to-4d dynamic scene generation

    Uriel Singer, Shelly Sheynin, Adam Polyak, Oron Ashual, Iurii Makarov, Filippos Kokkinos, Naman Goyal, Andrea Vedaldi, Devi Parikh, Justin Johnson, et al. Text-to-4d dynamic scene generation. In International Conference on Machine Learning (ICML), 2023

    2023

  44. [53]

    Energetically consistent invertible elasticity

    Alexey Stomakhin, Russell Howes, Craig A Schroeder, and Joseph M Teran. Energetically consistent invertible elasticity. In Symposium on Computer Animation, 2012

    2012

  45. [54]

    A material point method for snow simulation

    Alexey Stomakhin, Craig Schroeder, Lawrence Chai, Joseph Teran, and Andrew Selle. A material point method for snow simulation. ACM Transactions on Graphics (TOG), 2013

    2013

  46. [55]

    Augmented mpm for phase-change and varied materials

    Alexey Stomakhin, Craig Schroeder, Chenfanfu Jiang, Lawrence Chai, Joseph Teran, and Andrew Selle. Augmented mpm for phase-change and varied materials. ACM Transactions on Graphics (TOG), 2014

    2014

  47. [56]

    Dreamgaussian: Generative gaussian splatting for efficient 3d content creation

    Jiaxiang Tang, Jiawei Ren, Hang Zhou, Ziwei Liu, and Gang Zeng. Dreamgaussian: Generative gaussian splatting for efficient 3d content creation. arXiv preprint arXiv:2309.16653, 2023

    2023 arXiv

  48. [57]

    Modelscope text-to-video technical report

    Jiuniu Wang, Hangjie Yuan, Dayou Chen, Yingya Zhang, Xiang Wang, and Shiwei Zhang. Modelscope text-to-video technical report. arXiv preprint arXiv:2308.06571, 2023

    2023 arXiv

  49. [58]

    Vidu4d: Single generated video to high-fidelity 4d reconstruction with dynamic gaussian surfels

    Yikai Wang, Xinzhou Wang, Zilong Chen, Zhengyi Wang, Fuchun Sun, and Jun Zhu. Vidu4d: Single generated video to high-fidelity 4d reconstruction with dynamic gaussian surfels. arXiv preprint arXiv:2405.16822, 2024 a

    2024 arXiv

  50. [59]

    Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation

    Zhengyi Wang, Cheng Lu, Yikai Wang, Fan Bao, Chongxuan Li, Hang Su, and Jun Zhu. Prolificdreamer: High-fidelity and diverse text-to-3d generation with variational score distillation. Advances in Neural Information Processing Systems (NeurIPS), 2024 b

    2024

  51. [60]

    Image quality assessment: from error visibility to structural similarity

    Zhou Wang, Alan C Bovik, Hamid R Sheikh, and Eero P Simoncelli. Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 2004

    2004

  52. [61]

    Godiva: Generating open-domain videos from natural descriptions

    Chenfei Wu, Lun Huang, Qianxi Zhang, Binyang Li, Lei Ji, Fan Yang, Guillermo Sapiro, and Nan Duan. Godiva: Generating open-domain videos from natural descriptions. arXiv preprint arXiv:2104.14806, 2021

    2021 arXiv

  53. [62]

    Physgaussian: Physics-integrated 3d gaussians for generative dynamics

    Tianyi Xie, Zeshun Zong, Yuxing Qiu, Xuan Li, Yutao Feng, Yin Yang, and Chenfanfu Jiang. Physgaussian: Physics-integrated 3d gaussians for generative dynamics. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    2024

  54. [63]

    Diffusion probabilistic modeling for video generation

    Ruihan Yang, Prakhar Srivastava, and Stephan Mandt. Diffusion probabilistic modeling for video generation. Entropy, 2023

    2023

  55. [64]

    Dragnuwa: Fine-grained control in video generation by integrating text, image, and trajectory

    Shengming Yin, Chenfei Wu, Jian Liang, Jie Shi, Houqiang Li, Gong Ming, and Nan Duan. Dragnuwa: Fine-grained control in video generation by integrating text, image, and trajectory. arXiv preprint arXiv:2308.08089, 2023 a

    2023 arXiv

  56. [65]

    4dgen: Grounded 4d content generation with spatial-temporal consistency

    Yuyang Yin, Dejia Xu, Zhangyang Wang, Yao Zhao, and Yunchao Wei. 4dgen: Grounded 4d content generation with spatial-temporal consistency. arXiv preprint arXiv:2312.17225, 2023 b

    2023 arXiv

  57. [66]

    Point-bert: Pre-training 3d point cloud transformers with masked point modeling

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition (CVPR), 2022

    2022

  58. [67]

    Hybrid grains: Adaptive coupling of discrete and continuum simulations of granular media

    Yonghao Yue, Breannan Smith, Peter Yichen Chen, Maytee Chantharayukhonthorn, Ken Kamrin, and Eitan Grinspun. Hybrid grains: Adaptive coupling of discrete and continuum simulations of granular media. ACM Transactions on Graphics (TOG), 2018

    2018

  59. [68]

    Stag4d: Spatial-temporal anchored generative 4d gaussians

    Yifei Zeng, Yanqin Jiang, Siyu Zhu, Yuanxun Lu, Youtian Lin, Hao Zhu, Weiming Hu, Xun Cao, and Yao Yao. Stag4d: Spatial-temporal anchored generative 4d gaussians. In European Conference on Computer Vision (ECCV), 2024

    2024

  60. [69]

    4diffusion: Multi-view video diffusion model for 4d generation

    Haiyu Zhang, Xinyuan Chen, Yaohui Wang, Xihui Liu, Yunhong Wang, and Yu Qiao. 4diffusion: Multi-view video diffusion model for 4d generation. arXiv preprint arXiv:2405.20674, 2024 a

    2024 arXiv

  61. [70]

    I2vgen-xl: High-quality image-to-video synthesis via cascaded diffusion models

    Shiwei Zhang, Jiayu Wang, Yingya Zhang, Kang Zhao, Hangjie Yuan, Zhiwu Qin, Xiang Wang, Deli Zhao, and Jingren Zhou. I2vgen-xl: High-quality image-to-video synthesis via cascaded diffusion models. arXiv preprint arXiv:2311.04145, 2023

    2023 arXiv

  62. [71]

    Physdreamer: Physics-based interaction with 3d objects via video generation

    Tianyuan Zhang, Hong-Xing Yu, Rundi Wu, Brandon Y Feng, Changxi Zheng, Noah Snavely, Jiajun Wu, and William T Freeman. Physdreamer: Physics-based interaction with 3d objects via video generation. In European Conference on Computer Vision (ECCV), 2024 b

    2024

  63. [72]

    Animate124: Animating one image to 4d dynamic scene

    Yuyang Zhao, Zhiwen Yan, Enze Xie, Lanqing Hong, Zhenguo Li, and Gim Hee Lee. Animate124: Animating one image to 4d dynamic scene. arXiv preprint arXiv:2311.14603, 2023

    2023 arXiv

  64. [73]

    A unified approach for text-and image-guided 4d scene generation

    Yufeng Zheng, Xueting Li, Koki Nagano, Sifei Liu, Otmar Hilliges, and Shalini De Mello. A unified approach for text-and image-guided 4d scene generation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024

    2024

  65. [74]

    Reconstruction and simulation of elastic objects with spring-mass 3d gaussians

    Licheng Zhong, Hong-Xing Yu, Jiajun Wu, and Yunzhu Li. Reconstruction and simulation of elastic objects with spring-mass 3d gaussians. European Conference on Computer Vision (ECCV), 2024

    2024

  66. [75]

    Neural stress fields for reduced-order elastoplasticity and fracture

    Zeshun Zong, Xuan Li, Minchen Li, Maurizio M Chiaramonte, Wojciech Matusik, Eitan Grinspun, Kevin Carlberg, Chenfanfu Jiang, and Peter Yichen Chen. Neural stress fields for reduced-order elastoplasticity and fracture. In SIGGRAPH Asia 2023 Conference Papers, 2023

    2023

  67. [76]

    @esa (Ref

    \@ifxundefined[1] #1\@undefined \@firstoftwo \@secondoftwo \@ifnum[1] #1 \@firstoftwo \@secondoftwo \@ifx[1] #1 \@firstoftwo \@secondoftwo [2] @ #1 \@temptokena #2 #1 @ \@temptokena \@ifclassloaded agu2001 natbib The agu2001 class already includes natbib coding, so you should ...

  68. [77]

    \@lbibitem[] @bibitem@first@sw\@secondoftwo \@lbibitem[#1]#2 \@extra@b@citeb \@ifundefined br@#2\@extra@b@citeb \@namedef br@#2 \@nameuse br@#2\@extra@b@citeb \@ifundefined b@#2\@extra@b@citeb @num @parse #2 @tmp #1 NAT@b@open@#2 NAT@b@shut@#2 \@ifnum @merge>\@ne @bibitem@firs...

  69. [78]

    o rper im plastisch-deformablen Zustand , author= Mises, R v , journal= Nachrichten von der Gesellschaft der Wissenschaften zu G \

    @open @close @open @close and [1] URL: #1 \@ifundefined chapter * \@mkboth \@ifxundefined @sectionbib * \@mkboth * \@mkboth\@gobbletwo \@ifclassloaded amsart * \@ifclassloaded amsbook * \@ifxundefined @heading @heading NAT@ctr thebibliography [1] @ \@biblabel @NAT@ctr \@bibset...

    2019 arXiv

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.