pith. machine review for the scientific record. sign in

arxiv: 2604.02799 · v1 · submitted 2026-04-03 · 💻 cs.CV

Recognition: no theorem link

UNICA: A Unified Neural Framework for Controllable 3D Avatars

Jiahe Zhu , Xinyao Wang , Yiyu Zhuang , Yanwen Wang , Jing Tian , Yao Yao , Hao Zhu
Authors on Pith no claims yet

Pith reviewed 2026-05-13 19:43 UTC · model grok-4.3

classification 💻 cs.CV
keywords controllable 3D avatarsneural generative modelaction-conditioned diffusionGaussian Splattingmotion planningriggingphysical simulation3D rendering
0
0 comments X

The pith

UNICA unifies motion planning, rigging, physical simulation, and rendering into one skeleton-free neural model for controllable 3D avatars from keyboard inputs.

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

The paper introduces UNICA as a generative model that accepts game-style keyboard controls and produces the next frame of a 3D avatar's geometry via an action-conditioned diffusion process on 2D position maps. A point transformer then converts the geometry into 3D Gaussian Splatting primitives for high-fidelity rendering. This single framework replaces the traditional multi-stage pipeline of appearance modeling, motion planning, rigging, and explicit physical simulation. A sympathetic reader would care because the method generates natural dynamics for hair and loose clothing while supporting long autoregressive sequences, potentially simplifying interactive avatar creation for games and virtual environments.

Core claim

UNICA is a skeleton-free generative model that unifies the workflow of motion planning, rigging, physical simulation, and rendering by generating avatar geometry through an action-conditioned diffusion model operating on 2D position maps and then mapping the result to 3D Gaussian Splatting for free-view rendering.

What carries the argument

Action-conditioned diffusion model on 2D position maps, followed by a point transformer to 3D Gaussian Splatting primitives.

Load-bearing premise

An action-conditioned diffusion model operating on 2D position maps can reliably capture complex dynamics such as hair and loose clothing without any explicit physical simulation or skeleton.

What would settle it

Generate long sequences with rapid movements of loose clothing or long hair and check whether the resulting deformations match observed real-world video dynamics or deviate in ways that break physical plausibility.

Figures

Figures reproduced from arXiv: 2604.02799 by Hao Zhu, Jiahe Zhu, Jing Tian, Xinyao Wang, Yanwen Wang, Yao Yao, Yiyu Zhuang.

Figure 1
Figure 1. Figure 1: UNICA is a unified model that generates action-controlled, 360 ◦ -renderable 3D avatars with dynamics. For the first time, UNICA unifies a workflow of "motion planning, rigging, physical simulation, and rendering" within a single model. Abstract. Controllable 3D human avatars have found widespread ap￾plications in 3D games, the metaverse, and AR/VR scenarios. The con￾ventional approach to creating such a 3… view at source ↗
Figure 1
Figure 1. Figure 1: To the best of our knowledge, this work is the first to explore end-to-end [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: The pipeline of UNICA. UNICA consists of an action-conditioned multi-frame diffusion model for avatar geometry and a point transformer for point-to-3DGS ap￾pearance mapping. The diffusion model takes latents of three position maps as context and generates one subsequent position map conditioned on a chosen action embedding. The generated position map is upscaled using bilinear interpolation and combined wi… view at source ↗
Figure 3
Figure 3. Figure 3: The position map rendering process and visualization of a four-frame group. (a) We use an A-Pose mesh of the avatar as geometry and the vertex coordinates of the posed avatar as vertex colors to render position maps. The position maps are rendered from six orthogonal views. (b) We partition the motion sequence into groups of four frames and normalize each group. Group Normalization. A practical challenge a… view at source ↗
Figure 4
Figure 4. Figure 4: Demonstration of progressive 4D inference. During autoregressive inference of UNICA, each round generates a relative movement that is accumulated in 3D space for the actual movement of the 3D avatar. The output frame of round n will be renor￾malized before it is used as input for round n+1 . Component Analysis (PCA) fitted on the training set to reconstruct the gener￾ated position map, aligning the six vie… view at source ↗
Figure 5
Figure 5. Figure 5: Animation results of UNICA demonstrating avatar response to key presses. For visualization clarity, we sample one frame every three frames along the trajectory [PITH_FULL_IMAGE:figures/full_fig_p011_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: FVD scores over a 2000-frame autoregressively generated sequence, computed using a sliding window of 200 frames with a stride of 10. The red dashed line indicates the linear trend, and the gray shaded region denotes ±1 standard deviation. The near￾flat trend demonstrates stable generation quality over extended rollouts. 4 Experiment 4.1 Animation Results Implementation Details. All models are trained on 8 … view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative comparison between UNICA and baseline methods across different frames and viewpoints. the mean without upward drift, confirming that UNICA maintains stable quality over extended rollouts. We further evaluate control responsiveness: the generated sequences require on average 13.4 and 9.8 frames for 180◦ and 90◦ turns, respec￾tively, closely matching the ground-truth values of 12 and 9. Given tha… view at source ↗
Figure 8
Figure 8. Figure 8: Comparison on the number of context frames. Four consecutive frames are shown for each method. ity. UNICA produces high-quality avatars closely matching ground truth, with strong performance on faces, bodies, and textural details, while generating nat￾ural dynamics for loose clothing. While Mixamo achieves high rendering quality through ground-truth mesh and texture mapping, automatic rigging fails to mode… view at source ↗
Figure 9
Figure 9. Figure 9: Ablation study of position map alignment via PCA reconstruction. Position Map Alignment. Each generated frame of avatar geometry com￾prises position maps from six views. Although spatial attention provides global reasoning for multi-view consistency, generation inaccuracies can cause misalign￾ment. As shown in [PITH_FULL_IMAGE:figures/full_fig_p014_9.png] view at source ↗
read the original abstract

Controllable 3D human avatars have found widespread applications in 3D games, the metaverse, and AR/VR scenarios. The conventional approach to creating such a 3D avatar requires a lengthy, intricate pipeline encompassing appearance modeling, motion planning, rigging, and physical simulation. In this paper, we introduce UNICA (UNIfied neural Controllable Avatar), a skeleton-free generative model that unifies all avatar control components into a single neural framework. Given keyboard inputs akin to video game controls, UNICA generates the next frame of a 3D avatar's geometry through an action-conditioned diffusion model operating on 2D position maps. A point transformer then maps the resulting geometry to 3D Gaussian Splatting for high-fidelity free-view rendering. Our approach naturally captures hair and loose clothing dynamics without manually designed physical simulation, and supports extra-long autoregressive generation. To the best of our knowledge, UNICA is the first model to unify the workflow of "motion planning, rigging, physical simulation, and rendering". Code is released at https://github.com/zjh21/UNICA.

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.

Referee Report

3 major / 1 minor

Summary. The paper introduces UNICA, a skeleton-free generative model for controllable 3D avatars. Given keyboard inputs, an action-conditioned diffusion model generates the next frame of avatar geometry via 2D position maps; a point transformer then converts this geometry to 3D Gaussian Splatting for free-view rendering. The central claim is that this single neural framework unifies motion planning, rigging, physical simulation, and rendering, naturally capturing non-rigid dynamics such as hair and loose clothing without explicit physics or skeletons, while supporting extra-long autoregressive rollouts.

Significance. If the implicit dynamics capture and unification hold under quantitative scrutiny, the work could simplify avatar pipelines for games, metaverse, and AR/VR applications by removing multi-stage engineering. The public code release is a clear strength for reproducibility. However, the absence of reported error metrics, ablations, or comparisons in the abstract leaves the practical advantage over conventional pipelines unverified.

major comments (3)
  1. [Abstract] Abstract: the claim that the diffusion model 'naturally captures hair and loose clothing dynamics without manually designed physical simulation' is load-bearing for the unification thesis, yet no quantitative results, ablation studies, or error analysis are provided to demonstrate that 2D position maps suffice for 3D non-rigid motion without drift or loss of depth ordering.
  2. [Abstract] Abstract: the assertion that UNICA is 'the first model to unify the workflow of motion planning, rigging, physical simulation, and rendering' requires a concrete comparison section showing how prior skeleton-based or hybrid methods fail at end-to-end unification; without such evidence the novelty claim remains ungrounded.
  3. [Abstract] The autoregressive generation claim for 'extra-long' sequences rests on the diffusion model avoiding geometric drift, but the 2D map representation and lack of explicit skeleton or physics prior create a high risk of accumulation errors over long horizons; this must be tested with metrics on sequence length and out-of-distribution actions.
minor comments (1)
  1. [Abstract] The abstract states 'Code is released at https://github.com/zjh21/UNICA' but provides no details on the exact release contents (e.g., training scripts, pretrained weights, or evaluation protocols).

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive feedback on our abstract claims. We agree that strengthening the grounding of our assertions with explicit references to experimental evidence will improve the manuscript. We address each major comment point by point below and outline the revisions we will make.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that the diffusion model 'naturally captures hair and loose clothing dynamics without manually designed physical simulation' is load-bearing for the unification thesis, yet no quantitative results, ablation studies, or error analysis are provided to demonstrate that 2D position maps suffice for 3D non-rigid motion without drift or loss of depth ordering.

    Authors: We acknowledge that the abstract would benefit from direct references to supporting results. The full paper presents qualitative results in Section 4.2 and quantitative geometry error metrics (e.g., Chamfer distance and normal consistency) in Section 5.1 that demonstrate effective capture of non-rigid dynamics such as hair and loose clothing. In the revision we will update the abstract to cite these metrics briefly and add a dedicated ablation on the 2D position map representation to quantify its handling of depth ordering and drift. revision: partial

  2. Referee: [Abstract] Abstract: the assertion that UNICA is 'the first model to unify the workflow of motion planning, rigging, physical simulation, and rendering' requires a concrete comparison section showing how prior skeleton-based or hybrid methods fail at end-to-end unification; without such evidence the novelty claim remains ungrounded.

    Authors: We agree that a dedicated comparison would better substantiate the unification claim. In the revised manuscript we will insert a new subsection (likely in Related Work or Experiments) that includes a comparison table and analysis contrasting UNICA against representative skeleton-based and hybrid pipelines, explicitly showing the multi-stage engineering required by prior methods versus our single neural framework. revision: yes

  3. Referee: [Abstract] The autoregressive generation claim for 'extra-long' sequences rests on the diffusion model avoiding geometric drift, but the 2D map representation and lack of explicit skeleton or physics prior create a high risk of accumulation errors over long horizons; this must be tested with metrics on sequence length and out-of-distribution actions.

    Authors: We recognize the importance of quantifying long-horizon stability. Our current experiments (Section 5.3 and supplementary material) evaluate autoregressive rollouts up to 1000 frames with reported per-frame error curves showing limited drift on in-distribution actions. To directly address the concern we will add explicit quantitative metrics on sequence length, error accumulation rates, and out-of-distribution action performance in the revised version. revision: partial

Circularity Check

0 steps flagged

No circularity: UNICA unification is an architectural design choice, not a derivation reducing to inputs

full rationale

The paper proposes a new generative model that combines an action-conditioned diffusion process on 2D position maps with a subsequent point transformer to produce 3D Gaussian Splatting output. This single-framework unification of motion planning, rigging, simulation, and rendering is presented as an empirical engineering result rather than a mathematical derivation. No equations are shown that define a quantity in terms of itself, no fitted parameters are relabeled as predictions, and no load-bearing claims rest on self-citations or imported uniqueness theorems. The central claim is therefore self-contained against external benchmarks and receives the default non-circularity finding.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The approach rests on standard generative-model assumptions rather than new axioms or invented entities. No free parameters are explicitly fitted in the abstract description.

axioms (1)
  • domain assumption Diffusion models conditioned on action signals can produce coherent next-frame 3D geometry from 2D position maps
    Invoked when the paper states the diffusion model generates geometry directly from keyboard inputs

pith-pipeline@v0.9.0 · 5509 in / 1290 out tokens · 36769 ms · 2026-05-13T19:43:38.693007+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

84 extracted references · 84 canonical work pages · 3 internal anchors

  1. [1]

    Adobe: Mixamo.https://www.mixamo.com/(2026), accessed: 2026-01-21

    work page 2026

  2. [2]

    In: CVPR

    Alldieck, T., Magnor, M., Bhatnagar, B.L., Theobalt, C., Pons-Moll, G.: Learning to reconstruct people in clothing from a single RGB camera. In: CVPR. pp. 1175– 1186 (2019)

    work page 2019

  3. [3]

    In: CVPR

    Bahmani, S., Skorokhodov, I., Rong, V., Wetzstein, G., Guibas, L., Wonka, P., Tulyakov, S., Park, J.J., Tagliasacchi, A., Lindell, D.B.: 4D-fy: Text-to-4D gener- ation using hybrid score distillation sampling. In: CVPR. pp. 7996–8006 (2024)

    work page 2024

  4. [4]

    https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/ (2025)

    Ball, P.J., Bauer, J., Belletti, F., et al.: Genie 3: A new frontier for world models. https://deepmind.google/blog/genie-3-a-new-frontier-for-world-models/ (2025)

    work page 2025

  5. [5]

    In: ECCV

    Baradel, F., Armando, M., Galaaoui, S., Brégier, R., Weinzaepfel, P., Rogez, G., Lucas, T.: Multi-HMR: Multi-person whole-body human mesh recovery in a single shot. In: ECCV. pp. 202–218 (2024)

    work page 2024

  6. [6]

    Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets

    Blattmann, A., Dockhorn, T., Kulal, S., et al.: Stable video diffusion: Scaling latent video diffusion models to large datasets. arXiv preprint arXiv:2311.15127 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  7. [7]

    In: ICML

    Bruce, J., Dennis, M., Edwards, A., et al.: Genie: Generative interactive environ- ments. In: ICML. pp. 4603–4623 (2024)

    work page 2024

  8. [8]

    In: CVPR (2017)

    Carreira, J., Zisserman, A.: Quo vadis, action recognition? a new model and the kinetics dataset. In: CVPR (2017)

    work page 2017

  9. [9]

    Chen, B., Martí Monsó, D., Du, Y., Simchowitz, M., Tedrake, R., Sitzmann, V.: DiffusionForcing:Next-tokenpredictionmeetsfull-sequencediffusion.In:NeurIPS. vol. 37, pp. 24081–24125 (2024)

    work page 2024

  10. [10]

    In: CVPR

    Chen, J., Hu, J., Wang, G., Jiang, Z., Zhou, T., Chen, Z., Lv, C.: TaoAvatar: Real-time lifelike full-body talking avatars for augmented reality via 3D Gaussian Splatting. In: CVPR. pp. 10723–10734 (2025)

    work page 2025

  11. [11]

    In: ICLR (2025)

    Chen, Y., Mihajlovic, M., Chen, X., Wang, Y., Prokudin, S., Tang, S.: Splatformer: Point transformer for robust 3D gaussian splatting. In: ICLR (2025)

    work page 2025

  12. [12]

    ACM Comput

    Ding, J., Zhang, Y., Shang, Y., et al.: Understanding world or predicting future? A comprehensive survey of world models. ACM Comput. Surv.58(3) (2025)

    work page 2025

  13. [13]

    In: ECCV

    Dong, J., Shuai, Q., Zhang, Y., Liu, X., Zhou, X., Bao, H.: Motion capture from internet videos. In: ECCV. pp. 210–227 (2020)

    work page 2020

  14. [14]

    In: ICML

    Esser, P., Kulal, S., Blattmann, A., et al.: Scaling rectified flow transformers for high-resolution image synthesis. In: ICML. pp. 12606–12633 (2024)

    work page 2024

  15. [15]

    Feng, Y., Choutas, V., Bolkart, T., Tzionas, D., Black, M.J.: Collaborative regres- sion of expressive bodies using moderation. In: 3DV. pp. 792–804 (2021)

    work page 2021

  16. [16]

    In: NeurIPS

    Gao, S., Yang, J., Chen, L., Chitta, K., Qiu, Y., Geiger, A., Zhang, J., Li, H.: Vista: A generalizable driving world model with high fidelity and versatile controllability. In: NeurIPS. vol. 37, pp. 91560–91596 (2024)

    work page 2024

  17. [17]

    Advancing open-source world models,

    Gao, Z., Wang, Q., Zeng, Y., et al.: Advancing open-source world models. arXiv preprint arXiv:2601.20540 (2026)

    work page arXiv 2026

  18. [18]

    In: CVPR

    Guo, C., Jiang, T., Chen, X., Song, J., Hilliges, O.: Vid2Avatar: 3D avatar re- construction from videos in the wild via self-supervised scene decomposition. In: CVPR. pp. 12858–12868 (2023)

    work page 2023

  19. [19]

    In: CVPR

    Guo, C., Zou, S., Zuo, X., Wang, S., Ji, W., Li, X., Cheng, L.: Generating diverse and natural 3D human motions from text. In: CVPR. pp. 5152–5161 (2022)

    work page 2022

  20. [20]

    ACM TOG39(4), 60:1–60:12 (2020) 16 J

    Harvey, F.G., Yurick, M., Nowrouzezahrai, D., Pal, C.: Robust motion in- betweening. ACM TOG39(4), 60:1–60:12 (2020) 16 J. Zhuet al

    work page 2020

  21. [21]

    Matrix-game 2.0: An open-source real-time and streaming interactive world model

    He, X., Peng, C., Liu, Z., et al.: Matrix-Game 2.0: An open-source real-time and streaming interactive world model. arXiv preprint arXiv:2508.13009 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  22. [22]

    In: ICCV

    Hernandez, A., Gall, J., Moreno-Noguer, F.: Human motion prediction via spatio- temporal inpainting. In: ICCV. pp. 7134–7143 (2019)

    work page 2019

  23. [23]

    In: NeurIPS

    Ho, J., Jain, A., Abbeel, P.: Denoising diffusion probabilistic models. In: NeurIPS. vol. 33, pp. 6840–6851 (2020)

    work page 2020

  24. [24]

    In: NeurIPS Workshop (2021)

    Ho, J., Salimans, T.: Classifier-free diffusion guidance. In: NeurIPS Workshop (2021)

    work page 2021

  25. [25]

    ACM TOG36(4), 42:1–42:13 (2017)

    Holden, D., Komura, T., Saito, J.: Phase-functioned neural networks for character control. ACM TOG36(4), 42:1–42:13 (2017)

    work page 2017

  26. [26]

    In: CVPR

    Hu, L., Zhang, H., Zhang, Y., Zhou, B., Liu, B., Zhang, S., Nie, L.: GaussianAvatar: Towards realistic human avatar modeling from a single video via animatable 3D Gaussians. In: CVPR. pp. 634–644 (2024)

    work page 2024

  27. [27]

    In: ICLR (2024)

    Jiang, Y., Zhang, L., Gao, J., Hu, W., Yao, Y.: Consistent4D: Consistent 360° dynamic object generation from monocular video. In: ICLR (2024)

    work page 2024

  28. [28]

    In: ICME

    Jiang, Y., Liao, Q., Wang, Z., Lin, X., Lu, Z., Zhao, Y., Wei, H., Ye, J., Zhang, Y., Shao, Z.: SMPLX-Lite: A realistic and drivable avatar benchmark with rich geometry and texture annotations. In: ICME. pp. 1–6 (2024)

    work page 2024

  29. [29]

    Nature638(8051), 656–663 (2025)

    Kanervisto, A., Bignell, D., Wen, L.Y., et al.: World and human action models towards gameplay ideation. Nature638(8051), 656–663 (2025)

    work page 2025

  30. [30]

    Kaufmann,M.,Aksan,E.,Song,J.,Pece,F.,Ziegler,R.,Hilliges,O.:Convolutional autoencoders for human motion infilling. In: 3DV. pp. 918–927 (2020)

    work page 2020

  31. [31]

    ACM TOG42(4), 139:1–139:14 (2023)

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3D Gaussian splatting for real-time radiance field rendering. ACM TOG42(4), 139:1–139:14 (2023)

    work page 2023

  32. [32]

    In: ICLR (2013)

    Kingma, D.P., Welling, M.: Auto-encoding variational bayes. In: ICLR (2013)

    work page 2013

  33. [33]

    ACM TOG40(4), 130:1–130:15 (2021)

    Li, P., Aberman, K., Hanocka, R., Liu, L., Sorkine-Hornung, O., Chen, B.: Learning skeletal articulations with neural blend shapes. ACM TOG40(4), 130:1–130:15 (2021)

    work page 2021

  34. [34]

    In: CVPR

    Li, Z., Zheng, Z., Wang, L., Liu, Y.: Animatable Gaussians: Learning pose- dependent Gaussian maps for high-fidelity human avatar modeling. In: CVPR. pp. 19711–19722 (2024)

    work page 2024

  35. [35]

    In: CVPR

    Lin, J., Zeng, A., Wang, H., Zhang, L., Li, Y.: One-stage 3D whole-body mesh recovery with component aware transformer. In: CVPR. pp. 21159–21168 (2023)

    work page 2023

  36. [36]

    ACM TOG44(4), 122:1–122:12 (2025)

    Liu, I., Xu, Z., Yifan, W., Tan, H., Xu, Z., Wang, X., Su, H., Shi, Z.: RigAny- thing: Template-free autoregressive rigging for diverse 3D assets. ACM TOG44(4), 122:1–122:12 (2025)

    work page 2025

  37. [37]

    ACM TOG34(6), 248:1–248:16 (2015)

    Loper, M., Mahmood, N., Romero, J., Pons-Moll, G., Black, M.J.: SMPL: A skinned multi-person linear model. ACM TOG34(6), 248:1–248:16 (2015)

    work page 2015

  38. [38]

    In: ICCV

    Mahmood, N., Ghorbani, N., Troje, N.F., Pons-Moll, G., Black, M.J.: AMASS: Archive of motion capture as surface shapes. In: ICCV. pp. 5442–5451 (2019)

    work page 2019

  39. [39]

    Yume-1.5: A text-controlled interactive world generation model,

    Mao, X., Li, Z., Li, C., Xu, X., Ying, K., He, T., Pang, J., Qiao, Y., Zhang, K.: Yume-1.5: A text-controlled interactive world generation model. arXiv preprint arXiv:2512.22096 (2025)

    work page arXiv 2025

  40. [40]

    In: CVPR

    Martinez, J., Black, M.J., Romero, J.: On human motion prediction using recurrent neural networks. In: CVPR. pp. 2891–2900 (2017)

    work page 2017

  41. [41]

    In: CVPR

    Mosella-Montoro, A., Ruiz-Hidalgo, J.: SkinningNet: Two-stream graph convolu- tional neural network for skinning prediction of synthetic characters. In: CVPR. pp. 18593–18602 (2022)

    work page 2022

  42. [42]

    In: NeurIPS

    van den Oord, A., Vinyals, O., kavukcuoglu, k.: Neural discrete representation learning. In: NeurIPS. vol. 30, pp. 6309–6318 (2017) UNICA: A Unified Neural Framework for Controllable 3D Avatars 17

    work page 2017

  43. [43]

    Parker-Holder, J., Ball, P., Bruce, J., et al.: Genie 2: A large-scale foundation world model.https://deepmind.google/discover/blog/genie- 2- a- large- scale- foundation-world-model/(2024)

    work page 2024

  44. [44]

    In: CVPR

    Pavlakos, G., Choutas, V., Ghorbani, N., Bolkart, T., Osman, A.A.A., Tzionas, D., Black, M.J.: Expressive body capture: 3D hands, face, and body from a single image. In: CVPR. pp. 10975–10985 (2019)

    work page 2019

  45. [45]

    Big Data 4(4), 236–252 (2016)

    Plappert, M., Mandery, C., Asfour, T.: The kit motion-language dataset. Big Data 4(4), 236–252 (2016)

    work page 2016

  46. [46]

    In: ICLR (2023)

    Poole, B., Jain, A., Barron, J.T., Mildenhall, B.: DreamFusion: Text-to-3D using 2D diffusion. In: ICLR (2023)

    work page 2023

  47. [47]

    In: CVPR

    Pumarola, A., Corona, E., Pons-Moll, G., Moreno-Noguer, F.: D-NeRF: Neural radiance fields for dynamic scenes. In: CVPR. pp. 10318–10327 (2021)

    work page 2021

  48. [48]

    In: NeurIPS

    Ren, J., Xie, K., Mirzaei, A., et al.: L4GM: Large 4D Gaussian reconstruction model. In: NeurIPS. vol. 37, pp. 56828–56858 (2024)

    work page 2024

  49. [49]

    In: CVPR

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: CVPR. pp. 10684–10695 (2022)

    work page 2022

  50. [50]

    ACM TOG36(6), 245:1–245:17 (2017)

    Romero, J., Tzionas, D., Black, M.J.: Embodied hands: Modeling and capturing hands and bodies together. ACM TOG36(6), 245:1–245:17 (2017)

    work page 2017

  51. [51]

    In: MICCAI

    Ronneberger, O., Fischer, P., Brox, T.: U-Net: Convolutional networks for biomed- ical image segmentation. In: MICCAI. pp. 234–241 (2015)

    work page 2015

  52. [52]

    In: NeurIPS

    Saharia, C., Chan, W., Saxena, S., et al.: Photorealistic text-to-image diffusion models with deep language understanding. In: NeurIPS. vol. 35, pp. 36479–36494 (2022)

    work page 2022

  53. [53]

    In: ICLR (2024)

    Shi, Y., Wang, P., Ye, J., Mai, L., Li, K., Yang, X.: MVDream: Multi-view diffusion for 3D generation. In: ICLR (2024)

    work page 2024

  54. [54]

    In: ICML

    Singer, U., Sheynin, S., Polyak, A., et al.: Text-to-4D dynamic scene generation. In: ICML. pp. 31915–31929 (2023)

    work page 2023

  55. [55]

    In: ICLR (2021)

    Song, J., Meng, C., Ermon, S.: Denoising diffusion implicit models. In: ICLR (2021)

    work page 2021

  56. [56]

    ACM TOG41(4), 136:1–136:13 (2022)

    Starke, S., Mason, I., Komura, T.: DeepPhase: Periodic autoencoders for learning motion phase manifolds. ACM TOG41(4), 136:1–136:13 (2022)

    work page 2022

  57. [57]

    ACM TOG38(6), 209:1–209:14 (2019)

    Starke, S., Zhang, H., Komura, T., Saito, J.: Neural state machine for character- scene interactions. ACM TOG38(6), 209:1–209:14 (2019)

    work page 2019

  58. [58]

    ACM TOG39(4), 54:1–54:13 (2020)

    Starke, S., Zhao, Y., Komura, T., Zaman, K.: Local motion phases for learning multi-contact character movements. ACM TOG39(4), 54:1–54:13 (2020)

    work page 2020

  59. [59]

    In: NeurIPS

    Su, S.Y., Yu, F., Zollhöfer, M., Rhodin, H.: A-NeRF: Articulated neural radiance fields for learning human shape, appearance, and pose. In: NeurIPS. vol. 34, pp. 12278–12291 (2021)

    work page 2021

  60. [60]

    Worldplay: Towards long-term geometric consistency for real-time interactive world modeling.arXiv preprint arXiv:2512.14614,

    Sun, W., Zhang, H., Wang, H., Wu, J., Wang, Z., Wang, Z., Wang, Y., Zhang, J., Wang, T., Guo, C.: WorldPlay: Towards long-term geometric consistency for real-time interactive world modeling. arXiv preprint arXiv:2512.14614 (2025)

    work page arXiv 2025

  61. [61]

    In: SIGGRAPH Asia

    Tao, H., Hou, S., Zou, C., Bao, H., Xu, W.: Neural motion graph. In: SIGGRAPH Asia. pp. 84:1–84:11 (2023)

    work page 2023

  62. [62]

    In: ICLR (2023)

    Tevet, G., Raab, S., Gordon, B., Shafir, Y., Cohen-Or, D., Bermano, A.H.: Human motion diffusion model. In: ICLR (2023)

    work page 2023

  63. [63]

    Towards Accurate Generative Models of Video: A New Metric & Challenges

    Unterthiner, T., Van Steenkiste, S., Kurach, K., Marinier, R., Michalski, M., Gelly, S.: Towards accurate generative models of video: A new metric & challenges. arXiv preprint arXiv:1812.01717 (2018)

    work page internal anchor Pith review Pith/arXiv arXiv 2018

  64. [64]

    In: ICLR (2025) 18 J

    Valevski, D., Leviathan, Y., Arar, M., Fruchter, S.: Diffusion models are real-time game engines. In: ICLR (2025) 18 J. Zhuet al

    work page 2025

  65. [65]

    In: ECCV

    Voleti, V., Yao, C.H., Boss, M., Letts, A., Pankratz, D., Tochilkin, D., Laforte, C., Rombach, R., Jampani, V.: SV3D: Novel multi-view synthesis and 3D generation from a single image using latent video diffusion. In: ECCV. pp. 439–457 (2024)

    work page 2024

  66. [66]

    In: SIGGRAPH

    Wang, L., Zhao, X., Sun, J., Zhang, Y., Zhang, H., Yu, T., Liu, Y.: StyleAvatar: Real-time photo-realistic portrait avatar from a single video. In: SIGGRAPH. pp. 67:1–67:10 (2023)

    work page 2023

  67. [67]

    arXiv preprint arXiv:2305.10874 (2023)

    Wang, W., Yang, H., Tuo, Z., He, H., Zhu, J., Fu, J., Liu, J.: VideoFactory: Swap attention in spatiotemporal diffusions for text-to-video generation. arXiv preprint arXiv:2305.10874 (2023)

    work page arXiv 2023

  68. [68]

    IEEE TIP13(4), 600–612 (2004)

    Wang, Z., Bovik, A.C., Sheikh, H.R., Simoncelli, E.P.: Image quality assessment: from error visibility to structural similarity. IEEE TIP13(4), 600–612 (2004)

    work page 2004

  69. [69]

    Wu, X., Jiang, L., Wang, P.S., Liu, Z., Liu, X., Qiao, Y., Ouyang, W., He, T., Zhao, H.:PointTransformerV3:Simplerfasterstronger.In:CVPR.pp.4840–4851(2024)

    work page 2024

  70. [70]

    In: ICLR (2025)

    Xie, Y., Yao, C.H., Voleti, V., Jiang, H., Jampani, V.: SV4D: Dynamic 3D content generation with multi-frame and multi-view consistency. In: ICLR (2025)

    work page 2025

  71. [71]

    Spatial-temporal transformer networks for traffic flow forecasting.arXiv preprint arXiv:2001.02908, 2020

    Xu, M., Dai, W., Liu, C., Gao, X., Lin, W., Qi, G.J., Xiong, H.: Spatial-temporal transformer networks for traffic flow forecasting. arXiv preprint arXiv:2001.02908 (2020)

    work page arXiv 2001

  72. [72]

    ACM TOG39(4), 58:1–58:14 (2020)

    Xu, Z., Zhou, Y., Kalogerakis, E., Landreth, C., Singh, K.: RigNet: Neural rigging for articulated characters. ACM TOG39(4), 58:1–58:14 (2020)

    work page 2020

  73. [73]

    Xu, Z., Zhou, Y., Kalogerakis, E., Singh, K.: Predicting animation skeletons for 3D articulated models via volumetric nets. In: 3DV. pp. 298–307 (2019)

    work page 2019

  74. [74]

    arXiv preprint arXiv:2503.13435 (2025)

    Yang, L., Zhu, K., Tian, J., Zeng, B., Lin, M., Pei, H., Zhang, W., Yan, S.: WideRange4D: Enabling high-quality 4D reconstruction with wide-range move- ments and scenes. arXiv preprint arXiv:2503.13435 (2025)

    work page arXiv 2025

  75. [75]

    In: ICLR (2025)

    Yang, Z., Pan, Z., Gu, C., Zhang, L.: Diffusion2: Dynamic 3D content generation via score composition of video and multi-view diffusion models. In: ICLR (2025)

    work page 2025

  76. [76]

    arXiv preprint arXiv:2503.16396 (2025)

    Yao, C.H., Xie, Y., Voleti, V., Jiang, H., Jampani, V.: SV4D 2.0: Enhancing spatio- temporal consistency in multi-view video diffusion for high-quality 4D generation. arXiv preprint arXiv:2503.16396 (2025)

    work page arXiv 2025

  77. [77]

    In: ICLR (2026)

    Yenphraphai, J., Mirzaei, A., Chen, J., Zou, J., Tulyakov, S., Yeh, R.A., Wonka, P., Wang, C.: ShapeGen4D: Towards high quality 4D shape generation from videos. In: ICLR (2026)

    work page 2026

  78. [78]

    ACM TOG37(4), 145:1–145:11 (2018)

    Zhang, H., Starke, S., Komura, T., Saito, J.: Mode-adaptive neural networks for quadruped motion control. ACM TOG37(4), 145:1–145:11 (2018)

    work page 2018

  79. [79]

    ACM TOG44(4), 123:1–123:18 (2025)

    Zhang, J.P., Pu, C.F., Guo, M.H., Cao, Y.P., Hu, S.M.: One model to rig them all: Diverse skeleton rigging with unirig. ACM TOG44(4), 123:1–123:18 (2025)

    work page 2025

  80. [80]

    In: CVPR

    Zhang, R., Isola, P., Efros, A.A., Shechtman, E., Wang, O.: The unreasonable effectiveness of deep features as a perceptual metric. In: CVPR. pp. 586–595 (2018)

    work page 2018

Showing first 80 references.