Pith. sign in

REVIEW 4 major objections 6 minor 47 references

PAMD: Plausibility-Aware Motion Diffusion Model for Long Dance Generation

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read PAMD, a plausibility-aware diffusion model, generates dances that are both musically aligned and physically realistic by constraining each pose with a learned neural distance field.

desk verdict A coherent integration of known components with consistent empirical gains, but the load-bearing plausibility field is underspecified, so the central claim is not yet verifiable. read the letter →

arxiv 2505.20056 v1 pith:4GEMGLBS submitted 2025-05-26 cs.CV

Hongsong Wang , Yin Zhu , Qiuxia Lai , Yang Zhang , Guo-Sen Xie , Xin Geng This is my paper · ORCID
classification cs.CV
keywords music-to-dancegenerationdiffusionmodelsneuraldistancefieldsmotionplausibilityfoot-groundcontactlongdancebeatalignmenthumanposemanifold
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

PAMD is a diffusion-based framework for music-to-dance generation that aims to fix the physical implausibility of generated dances—sliding feet, floating, joint penetration—by steering generated poses toward a learned manifold of real human poses. Its central claim is that a Neural Distance Field that scores each pose by its distance from plausible motion, used as an auxiliary loss during diffusion training, yields dances that are both better aligned to music beats and more physically realistic. The paper reports consistent wins over the EDGE baseline on beat alignment, foot-contact plausibility, and FID metrics, with the advantage growing on longer (7.5–10 s) sequences, and user studies prefer PAMD over both EDGE and ground-truth dances. The practical payoff would be computationally choreographed dances that look natural enough for VR, AR, and digital entertainment.

What carries the argument

The load-bearing object is the Neural Distance Field (NDF), a function $f: SO(3)^K \to \mathbb{R}_+$ whose zero level set is the manifold of plausible poses, with outputs above zero giving an unsigned distance from that manifold. It is built with an encoder that aggregates joint rotations up the kinematic tree and a decoder that predicts the distance; during diffusion training, the average distance of generated frames is added as the auxiliary loss $L_{PMC}$. The other two modules are auxiliary: Prior Motion Guidance (PMG) concatenates a fixed standing pose with music features and the timestep as the conditioning input to cross-attention, and the Motion Refinement with Foot-Ground Contact (MRFC) block computes foot positions, velocities, and contact labels, derives a contact score, and refines the raw dance in position space before the model's output is used.

What would settle it

A concrete test: train the full PAMD pipeline but replace the NDF with a randomly initialized network (same architecture, no training); if PFC and FID metrics do not worsen relative to the trained NDF, then the reported plausibility gains are not caused by the distance field. More directly, feed the NDF a set of physically impossible poses (e.g., feet penetrating the floor, knees hyperextended) that are nonetheless close to real poses in the latent space; if it assigns them near-zero distance, the field is not detecting the implausibility it is claimed to model.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that physical plausibility in diffusion-based dance generation can be enforced without a physics simulator: a network f trained to output zero on real, plausible poses and positive values elsewhere is used as a soft constraint. The training loss becomes $L = L_{recon} + \lambda_{joint} L_{joint} + \lambda_{vel} L_{vel} + \lambda_{foot} L_{foot} + \lambda_{PMC} L_{PMC}$, where $L_{PMC}$ is the average NDF output over the frames of the denoised dance, so the diffusion model is pushed toward the zero-level set of the plausibility field. Two further modules add a fixed standing pose as a conditioning prior (PMG) and a lightweight foot-contact refinement block (MRFC) that operates in joint-position space before converting back to rotations. The paper claims these components let PAMD generate long dances in parallel by overlapping sliding windows, without the error accumulation of autoregressive generation, and that the resulting motions have fewer skating, floating, and penetration artifacts.

Load-bearing premise

The load-bearing premise is that the NDF plausibility field, trained on real human poses, measures physical plausibility rather than mere proximity to the training distribution, so minimizing its output during diffusion training pulls generated motions toward physically valid poses.

Editorial extensions

If this is right

  • On the AIST++ benchmark, PAMD improves beat alignment score and physical foot contact over EDGE, and the gap widens as generation length grows from 7.5 s to 10 s.
  • Ablations show PMC, PMG, and MRFC are complementary: removing PMC raises PFC by roughly 29% (2.04 vs 1.44), while adding MRFC alone cuts PFC by about 6%.
  • The refinement module achieves comparable or better quality than Lodge's coarse-to-fine foot refine block at roughly one-third the total parameters (207.44M vs 804.72M) and a smaller refine module (1.50M vs 4.51M).
  • PAMD can generate long dances of arbitrary length in parallel by overlapping slices with a linear blend, avoiding the error accumulation of autoregressive inference.

Reading between the lines

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

  • The NDF field is scored per frame, so it likely cannot capture temporal physical plausibility such as momentum or ground reaction forces; a natural extension is a spatiotemporal distance field that scores short pose windows.
  • The paper does not describe the NDF's training set, split, or validation; a testable extension is to train the same PAMD pipeline with a distance field learned from a different motion corpus to see whether the plausibility gain transfers.
  • The sliding-window merge uses a linear ramp for blending; it would be worth testing whether the overlap length and blending function affect the beat-alignment improvements reported for long sequences.
  • If the NDF is essentially measuring proximity to the training distribution of poses, then its benefit may shrink on out-of-distribution music styles or unusual choreography; a stress test on novel genres would clarify whether the constraint generalizes or only memorizes.
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 / 6 minor

Summary. The paper proposes PAMD, a diffusion-based framework for music-to-dance generation on the AIST++ dataset. The method augments a standard motion diffusion model with three components: Plausible Motion Constraint (PMC), which uses a neural distance field to penalize implausible poses; Prior Motion Guidance (PMG), which appends a fixed standing pose as an auxiliary conditioning signal; and Motion Refinement with Foot-Ground Contact (MRFC), a lightweight cross-attention module for reducing foot-skating. Long sequences are generated by slicing the audio into overlapping 5-second segments, denoising them in parallel with latent-level replacement, and merging with weighted averaging. Experiments report improvements over EDGE on beat alignment, foot contact, FID, and user preference for 5s, 7.5s, and 10s dances, with ablations for each module.

Significance. If the central claim holds, PAMD is a useful contribution to long dance generation: it demonstrates a parallel generation scheme, a lightweight refinement module, and a new application of neural distance fields to music-conditioned motion synthesis. The paper provides an extensive evaluation with ablations, comparisons against EDGE and Lodge, and a user study. However, the physical-plausibility claim rests on the NDF's training and validity, which is not described, and the reported gains are not accompanied by variance or significance information. The novelty of applying NDFs to dance generation is real but the current manuscript does not substantiate it.

major comments (4)
  1. [§III-B, Eq. (7)] The Plausible Motion Constraint is the paper's core contribution, but the neural distance field f is never specified as a training objective, data split, or regularizer. The text gives the architecture (encoder/decoder MLPs with parent-joint aggregation and quaternion transform) and defines L_PMC = (1/N) Σ f_dec(f_enc(q(x̂_i))), but it does not state how f is trained, on which poses, with which loss, or whether it is frozen during end-to-end training. If f is trained jointly with the diffusion model without a distance regression objective, it can collapse to a near-constant small output and L_PMC becomes vacuous; if it is pretrained on the full AIST++ set, test-set improvements in Tables I–III could be inflated by memorization. Please provide the training protocol for the NDF, its validation (e.g., distance accuracy on held-out poses), and state whether it is frozen.
  2. [§I, §III-B, Table V] The central claim of 'physically realistic' motion is supported only by the learned NDF zero-level set of Eq. (4) and by the proxy metrics in Table V (Skating, floating, penetration), whose computation is not defined anywhere in the paper. A distance to a manifold learned from real motion data measures closeness to the training distribution, not physical validity; the paper provides no evidence (e.g., correlation with physics simulation or contact-force analysis) that the NDF distance corresponds to physical implausibility. Please define the Table V metrics precisely and show that minimizing the PMC loss reduces them in a controlled way, or provide external physical validation.
  3. [§III-E, Algorithm 1, Table II] The long-dance scheme is evaluated only for 7.5s and 10s outputs and against EDGE alone, which does not establish the claim of generating 'dances of arbitrary lengths'. More importantly, Algorithm 1 operates the overlap replacement on the noisy latent x_{t-1} at every reverse step (line 7), while the two slices are denoised from independent noise; this can create inconsistent latents at the boundary, and the final weighted averaging (lines 9–18) is not a substitute for enforcing motion coherence at the stitching points. Please quantify boundary continuity (e.g., velocity or acceleration discontinuity at slice boundaries) and evaluate longer sequences (e.g., 30s) with additional long-generation baselines.
  4. [§IV-A, Tables I–III] Evaluation metrics are reported as single numbers without standard deviations or significance tests. The implementation details state that the test process is repeated 100 times and averaged, but no variance is given; consequently, differences like BAS 0.31 vs. 0.30 in Table I or PFC 1.44 vs. 1.56 in Table III cannot be assessed for statistical significance. Please report error bars or run paired significance tests for the core comparisons against EDGE and for the ablations.
minor comments (6)
  1. [Figure 3] Figure 3 contains placeholder text ('?????') in the pipeline diagram, which should be replaced with the actual tensor/feature names.
  2. [§III-D, Eq. (9)] In Eq. (9), the variables h_max and v_max are used but never defined; please state how they are computed (presumably from dataset statistics).
  3. [§II-A] In Section II.A, the text claims the NDF 'constrains the motion to adhere to natural physical laws', but the motion representation is in rotation space and the NDF only models poses; this overstates what the module does.
  4. [Table I] In Table I, the TM2D method achieves Div_k = 9.10, closer to the ground truth 9.29 than PAMD's 5.94; the text should acknowledge this and discuss why lower diversity in the kinetic feature space is acceptable.
  5. [§IV-C] The user study in Section IV.C does not report how the 21 pairs were selected, whether the order was randomized, the expertise of the 11 participants, or inter-rater variability; please provide this information.
  6. [§IV-B] In Section IV.B, the phrase 'overcoming the poor quality and diversity of dances generated using Diffusion model in kinematic feature space' is incomplete; please clarify the intended comparison.

Circularity Check

2 steps flagged · score 4.0 of 10

PMC's plausibility is partly self-definitional: Eq. (4) defines “plausible” as the zero-level set of an NDF that, per Sec. III-B, models the “actual pose manifold” of real poses, and the abstract relabels that set as “physically valid”; the NDF's training (loss, data, frozen?) is undocumented, though independent metrics (PFC, artifact scores, user study) carry the empirical claims.

  1. self definitional [Abstract; Sec. III-B, Eqs. (4)-(7)]
    "Abstract: “...leverages Neural Distance Fields (NDFs) to model the actual pose manifold and guide generated motions toward a physically valid pose manifold.” Sec. III-B: “The manifold of plausible poses is represented as the zero-level set: S={θ∈SO(3)^K | f(θ)=0}, where SO(3)^K denotes the pose space... and the value of f(θ) signifies the unsigned distance from the pose θ to the manifold of plausibility.”"

    Eq. (4) defines plausibility as the zero-level set of f, and Sec. III-B says PMC “models the actual pose manifold, preserving distances between real poses,” so f is fit to real poses; a distance field's zero set is its training-pose set. The abstract then equates this statistical object with a “physically valid pose manifold.” Minimizing L_PMC (Eq. 7) pushes generated poses into the training-pose distribution by construction, so the claimed physical plausibility is, at this step, a relabeling of data-manifold matching: “plausible” means whatever poses f was trained on, and the outcome (physically valid poses) is contained in the input (real poses used to fit f). The NDF encodes no physical law; “adhere to natural physical laws” (Sec. II-A) is asserted, not derived.

  2. other [Sec. III-B, Eq. (7); Sec. IV-A Implementation Details]
    "Sec. III-B: “The PMC module consists of an encoder f_enc and a decoder f_dec, following the practice of modeling pose manifolds with neural distance fields [11].” Eq. (7): “L_PMC = 1/N Σ_i f_dec(f_enc(q(x̂_i))).” Sec. IV-A lists only batch size, epochs, learning rate and weight decay; no NDF training objective, data split, or freeze status."

    The NDF's training is never specified: no distance-supervision loss, no pose dataset, no train/test split, no statement of whether f is frozen when L_PMC is back-propagated. Eq. (7) is thus a sum of outputs of an unspecified network. If f is jointly trained without a distance objective it can collapse to a near-constant and L_PMC is vacuous; if f was fit to the same AIST++ poses used for evaluation, the FID/PFC gains in Tables I-III and V partly measure closeness to the memorized pose distribution rather than physical plausibility. This is a missing-support flag, not a deductive loop, but it is load-bearing: without the training specification, the “distance to the physically valid pose manifold” reading of Eq.

full rationale

Most of PAMD's empirical content is externally grounded: benchmarking on AIST++ against EDGE, Lodge, Bailando, TM2D, and BADM uses standard metrics (BAS, PFC, FID_k, FID_g, Div), and the physical measures in Table V (skating, floating, penetration) are computed from forward kinematics and contact labels, not from the learned NDF. MRFC and PMG are described with concrete objectives. So the method as a whole does not reduce to a single fit. The circular element is confined to PMC, which the abstract calls “the core of PAMD.” Eq. (4) defines “the manifold of plausible poses” as the zero-level set of a learned network f, and Sec. III-B says PMC “models the actual pose manifold, preserving distances between real poses.” Since f is fit to real human poses, its zero set lies where the training poses lie; the abstract's “physically valid pose manifold” and Sec. II-A's claim that the NDF “constrains the motion to adhere to natural physical laws” are relabelings of data-manifold membership. Minimizing L_PMC (Eq. 7) pulls generations toward the training pose manifold by construction. A missing-support problem compounds this: the paper never states how f_enc and f_dec are trained — no distance-supervision loss, no data split, no freeze status. If f is trained jointly without a distance objective, it can collapse and Eq. (7) becomes vacuous; if f was fit to the same AIST++ poses used for evaluation, the gains partly reflect distribution memorization. That is not itself a deductive loop, but it blocks independent grounding of the “physical plausibility” label. The only self-citation is ref. [45] (first author's PVRED) for the quaternion transform q(·) in Sec. III-B; it is standard mathematics and not load-bearing. Because PMC's plausibility notion is self-definitional and its training is undocumented, while headline results still rest on independent external metrics, the score is 4 rather than 6 or higher.

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

The central claim rests on a learned plausibility measure, a hand-picked prior pose, and a stitching heuristic. None of these are derived from first principles; they are inputs from prior work or modeling choices. The main free parameters are hyperparameters that are unreported or tuned by hand. The most consequential assumption is that the NDF distance field trained on real poses represents physical plausibility rather than data density.

free parameters (5)
  • Loss weights lambda_joint, lambda_vel, lambda_foot, lambda_PMC = not reported
    The total loss in Eq. (8) combines four auxiliary terms; the paper does not give values, and the balance between reconstruction and plausibility is tuned by hand.
  • Guidance weight w = 2 (default) and 1 for comparison
    Classifier-free guidance scale in Eq. (3); selected by hand; all main results are sensitive to this choice.
  • Foot-contact score scales k_h, k_v = k_h = 5*h_max, k_v = 5*v_max
    Constants in Eq. (9), taken from ProxyCap [46] rather than fitted to AIST++; they control how strongly height and velocity differences turn into contact scores.
  • Prior standing pose for PMG = a standard standing pose from the dataset
    The prior motion condition in Sec. III-C is a hand-picked pose; the ablation in Table VI shows performance varies with this choice.
  • Diffusion timesteps T = not reported
    Algorithm 1 uses T but the value is never stated; likely inherited from MDM/EDGE defaults.
assumptions (5)
  • standard math DDPM forward/reverse process and classifier-free guidance are valid generative machinery.
    Sec. III-A Eqs. (1)-(3) rely on DDPM and classifier-free guidance as unproved background.
  • domain assumption SMPL body model, 6D rotation representation, and forward kinematics faithfully describe human dance poses.
    Sec. III-A uses SMPL/6D representations and FK in auxiliary losses; standard in the field but not validated in this paper.
  • ad hoc to paper The learned NDF zero-level set is an accurate model of the physically plausible pose manifold, and distance to it is a valid plausibility measure.
    Sec. III-B Eqs. (4)-(7); the paper does not describe NDF training, data split, or validation, and equates data-manifold proximity with physical plausibility.
  • ad hoc to paper A fixed standard standing pose is a beneficial universal condition for all dances in the target distribution.
    Sec. III-C; the authors assert standing poses appear in transitions, beginnings, and endings of nearly all dances, without a distributional study.
  • ad hoc to paper Overlapping slicing with weighted averaging produces long sequences that preserve coherence and avoid boundary artifacts.
    Algorithm 1 and Sec. III-E; no analysis or experiment isolating stitching artifacts is provided.
invented entities (1)
  • Plausibility distance function f(theta) from the PMC Neural Distance Field
    purpose: Frame-level auxiliary loss that pushes generated poses toward the learned zero-level set of plausible poses.
    The function is trained on AIST++ poses and used both as a training signal and as evidence for plausibility; the paper provides no external validation that its outputs correspond to physical correctness.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PAMD: Plausibility-Aware Motion Diffusion Model for Long Dance Generation." pith.science (2026). https://pith.science/paper/4GEMGLBS

@misc{pith2026250520056,
  author       = {Pith},
  title        = {Pith review of: PAMD: Plausibility-Aware Motion Diffusion Model for Long Dance Generation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4GEMGLBS}},
  note         = {Machine review of arXiv:2505.20056}
}
read the original abstract

Computational dance generation is crucial in many areas, such as art, human-computer interaction, virtual reality, and digital entertainment, particularly for generating coherent and expressive long dance sequences. Diffusion-based music-to-dance generation has made significant progress, yet existing methods still struggle to produce physically plausible motions. To address this, we propose Plausibility-Aware Motion Diffusion (PAMD), a framework for generating dances that are both musically aligned and physically realistic. The core of PAMD lies in the Plausible Motion Constraint (PMC), which leverages Neural Distance Fields (NDFs) to model the actual pose manifold and guide generated motions toward a physically valid pose manifold. To provide more effective guidance during generation, we incorporate Prior Motion Guidance (PMG), which uses standing poses as auxiliary conditions alongside music features. To further enhance realism for complex movements, we introduce the Motion Refinement with Foot-ground Contact (MRFC) module, which addresses foot-skating artifacts by bridging the gap between the optimization objective in linear joint position space and the data representation in nonlinear rotation space. Extensive experiments show that PAMD significantly improves musical alignment and enhances the physical plausibility of generated motions. This project page is available at: https://mucunzhuzhu.github.io/PAMD-page/.

Figures

Figures reproduced from arXiv: 2505.20056 by the authors.

Figure 1
Figure 1. Motivation of our approach for diffusion-based music-to-dance generation. To generate plausible and correct motion sequences for music￾to-dance, we introduce prior motion and plausible motion constraints during the training of the generative diffusion model. Furthermore, computational music-to-dance generation fosters interdisciplinary collaboration among computer science, per￾forming arts, and human-computer intera… view at source ↗
Figure 2
Figure 2. Our PAMD model generates long dances that are better synchronized and visually coherent. The black dots on the pink music waveform indicate music beats, while the grey and blue motions denote dance beats generated by PAMD (ours) and EDGE, respectively. The underlined dance beats indicate close alignment, which falls within five frames of the nearest music beat. PAMD produces eight closely aligned dance beats compare… view at source ↗
Figure 3
Figure 3. PAMD Pipeline Overview: Conditioned on music and prior motion, PAMD learns to denoise dance sequences from time t = T to t = 0. Music features are extracted by Jukebox and then pass through the Transformer Music Encoder. The prior motion, timestep, and music features are concatenated and undergo cross-attention with noise. The noisy sequence xˆt is processed by a transformer-based dance decoder, which generates the … view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Implausible poses of dance generation: The score indicates the implausibility of dance poses, with higher scores indicating less plausible poses. It is predicted by the trained auxiliary network in the PMC module. where FK(·) denotes the forward kinematic that converts…
Figure 5
Figure 5. Figure 5: Prior Motion Guidance: xprior is the chosen prior motion; m and t are music features and timestep token; xˆt is the input noisy dance sequence. x˜0 denotes the output raw dance. Algorithm 1 Long dance generation algorithm Input: The conditioned music. Parameter: Diffus…
Figure 6
Figure 6. Figure 6 [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Visualization of Long Dance Generation: The dances with a duration of 10 seconds are shown in (a) break, (b) pop, and (c) waack. Each frame is sampled at 0.5-second intervals, resulting in 20 frames for each 10-second dance sequence [PITH_FULL_IMAGE:figures/full_fig_p…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

47 extracted references · 40 canonical work pages

  1. [1]

    Expression of emotions in dance: Relation between arm movement characteristics and emotion,

    M. Sawada, K. Suda, and M. Ishii, “Expression of emotions in dance: Relation between arm movement characteristics and emotion,”Percep- tual and Motor Skills, vol. 97, no. 3, pp. 697–708, 2003

    work page 2003

  2. [2]

    L. Georgios, “The transformation of traditional dance from its first to its second existence: The effectiveness of music-movement education and creative dance in the preservation of our cultural heritage.”Journal of Education and Training Studies, vol. 6, no. 1, pp. 104–112, 2018

    work page 2018

  3. [3]

    Evolution and functions of human dance,

    B. Fink, B. Bl ¨asing, A. Ravignani, and T. K. Shackelford, “Evolution and functions of human dance,”Evolution and Human Behavior, vol. 42, no. 4, pp. 351–360, 2021

    work page 2021

  4. [4]

    Edge: Editable dance generation from music,

    J. Tseng, R. Castellon, and K. Liu, “Edge: Editable dance generation from music,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2023, pp. 448–458

    work page 2023

  5. [5]

    Dancing-to-music character animation,

    T. Shiratori, A. Nakazawa, and K. Ikeuchi, “Dancing-to-music character animation,” inComputer Graphics Forum, vol. 25, no. 3, 2006, pp. 449– 458

    work page 2006

  6. [6]

    A game-like appli- cation for dance learning using a natural human computer interface,

    A. Kitsikidis, K. Dimitropoulos, D. U ˘gurca, C. Bayc ¸ay, E. Yilmaz, F. Tsalakanidou, S. Douka, and N. Grammalidis, “A game-like appli- cation for dance learning using a natural human computer interface,” inInternational Conference on Universal Access in Human-Computer Interaction, 2015, pp. 472–482

    work page 2015

  7. [7]

    Ai choreographer: Music conditioned 3d dance generation with aist++,

    R. Li, S. Yang, D. A. Ross, and A. Kanazawa, “Ai choreographer: Music conditioned 3d dance generation with aist++,” inIEEE/CVF International Conference on Computer Vision, 2021, pp. 13 401–13 412

    work page 2021

  8. [8]

    A brand new dance partner: Music-conditioned pluralistic dancing controlled by multiple dance genres,

    J. Kim, H. Oh, S. Kim, H. Tong, and S. Lee, “A brand new dance partner: Music-conditioned pluralistic dancing controlled by multiple dance genres,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 3490–3500. 9 (a) Break (b) Pop EDGEOursGTEDGEOursGTEDGEOursGT (c) Waack Fig. 7.Visualization of Long Dance Generation:The dances with...

    work page 2022

Show all 47 references
  1. [9]

    Bailando: 3d dance generation by actor-critic gpt with choreographic memory,

    S. Li, W. Yu, T. Gu, C. Lin, Q. Wang, C. Qian, C. C. Loy, and Z. Liu, “Bailando: 3d dance generation by actor-critic gpt with choreographic memory,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2022, pp. 11 050–11 059

    2022

  2. [10]

    Genre-conditioned long-term 3d dance generation driven by music,

    Y . Huang, J. Zhang, S. Liu, Q. Bao, D. Zeng, Z. Chen, and W. Liu, “Genre-conditioned long-term 3d dance generation driven by music,” inIEEE International Conference on Acoustics, Speech and Signal Processing. IEEE, 2022, pp. 4858–4862

    2022

  3. [11]

    Pose-ndf: Modeling human pose manifolds with neural distance fields,

    G. Tiwari, D. Anti ´c, J. E. Lenssen, N. Sarafianos, T. Tung, and G. Pons- Moll, “Pose-ndf: Modeling human pose manifolds with neural distance fields,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 572–589

    2022

  4. [12]

    Humor: 3d human motion model for robust pose estimation,

    D. Rempe, T. Birdal, A. Hertzmann, J. Yang, S. Sridhar, and L. J. Guibas, “Humor: 3d human motion model for robust pose estimation,” inIEEE/CVF International Conference on Computer Vision, 2021, pp. 11 488–11 499

    2021

  5. [13]

    A unified framework for real time motion completion,

    Y . Duan, Y . Lin, Z. Zou, Y . Yuan, Z. Qian, and B. Zhang, “A unified framework for real time motion completion,” inProceedings of the AAAI Conference on Artificial Intelligence, vol. 36, no. 4, 2022, pp. 4459– 4467

    2022

  6. [14]

    Lodge: A coarse to fine diffusion network for long dance generation guided by the characteristic dance primitives,

    R. Li, Y . Zhang, Y . Zhang, H. Zhang, J. Guo, Y . Zhang, Y . Liu, and X. Li, “Lodge: A coarse to fine diffusion network for long dance generation guided by the characteristic dance primitives,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognitio...

    2024

  7. [15]

    Learn2dance: Learning statistical music-to-dance mappings for choreography synthesis,

    F. Ofli, E. Erzin, Y . Yemez, and A. M. Tekalp, “Learn2dance: Learning statistical music-to-dance mappings for choreography synthesis,”IEEE Transactions on Multimedia, vol. 14, no. 3, pp. 747–759, 2011

    2011

  8. [16]

    Example-based automatic music-driven conventional dance motion synthesis,

    R. Fan, S. Xu, and W. Geng, “Example-based automatic music-driven conventional dance motion synthesis,”IEEE Transactions on Visualiza- tion and Computer Graphics, vol. 18, no. 3, pp. 501–515, 2011

    2011

  9. [17]

    Kinetic imaginations: Exploring the possi- bilities of combining ai and dance

    A. Berman and V . James, “Kinetic imaginations: Exploring the possi- bilities of combining ai and dance.” inIJCAI, 2015, p. 2431

    2015

  10. [18]

    Music similarity-based approach to generating dance motion sequence,

    M. Lee, K. Lee, and J. Park, “Music similarity-based approach to generating dance motion sequence,”Multimedia tools and applications, vol. 62, pp. 895–912, 2013

    2013

  11. [19]

    Transflower: probabilistic autoregressive dance generation with multimodal attention,

    G. Valle-P ´erez, G. E. Henter, J. Beskow, A. Holzapfel, P.-Y . Oudeyer, and S. Alexanderson, “Transflower: probabilistic autoregressive dance generation with multimodal attention,”ACM Transactions on Graphics (TOG), vol. 40, no. 6, pp. 1–14, 2021

    2021

  12. [20]

    Deepdance: music-to-dance motion choreography with adversarial learning,

    G. Sun, Y . Wong, Z. Cheng, M. S. Kankanhalli, W. Geng, and X. Li, “Deepdance: music-to-dance motion choreography with adversarial learning,”IEEE Transactions on Multimedia, vol. 23, pp. 497–509, 2020

    2020

  13. [21]

    Tm2d: Bimodality driven 3d dance generation via music- text integration,

    K. Gong, D. Lian, H. Chang, C. Guo, Z. Jiang, X. Zuo, M. B. Mi, and X. Wang, “Tm2d: Bimodality driven 3d dance generation via music- text integration,” inIEEE/CVF International Conference on Computer Vision, 2023, pp. 9942–9952

    2023

  14. [22]

    Bidirectional autoregessive diffusion model for dance generation,

    C. Zhang, Y . Tang, N. Zhang, R.-S. Lin, M. Han, J. Xiao, and S. Wang, “Bidirectional autoregessive diffusion model for dance generation,” in IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 687–696

    2024

  15. [23]

    Popdg: Popular 3d dance generation with popdanceset,

    Z. Luo, M. Ren, X. Hu, Y . Huang, and L. Yao, “Popdg: Popular 3d dance generation with popdanceset,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 26 984–26 993

    2024

  16. [24]

    Beat-it: Beat-synchronized multi-condition 3d dance generation,

    Z. Huang, X. Xu, C. Xu, H. Zhang, C. Zheng, J. Qin, and S. He, “Beat-it: Beat-synchronized multi-condition 3d dance generation,”arXiv preprint arXiv:2407.07554, 2024

    2024 arXiv

  17. [25]

    Dancecomposer: Dance-to- music generation using a progressive conditional music generator,

    X. Liang, W. Li, L. Huang, and C. Gao, “Dancecomposer: Dance-to- music generation using a progressive conditional music generator,”IEEE Transactions on Multimedia, 2024

    2024

  18. [26]

    Human motion generation: A survey,

    W. Zhu, X. Ma, D. Ro, H. Ci, J. Zhang, J. Shi, F. Gao, Q. Tian, and Y . Wang, “Human motion generation: A survey,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023

    2023

  19. [27]

    Action2motion: Conditioned generation of 3d human motions,

    C. Guo, X. Zuo, S. Wang, S. Zou, Q. Sun, A. Deng, M. Gong, and L. Cheng, “Action2motion: Conditioned generation of 3d human motions,” in28th ACM International Conference on Multimedia, 2020, pp. 2021–2029

    2020

  20. [28]

    Action-conditioned 3d human motion synthesis with transformer vae,

    M. Petrovich, M. J. Black, and G. Varol, “Action-conditioned 3d human motion synthesis with transformer vae,” inIEEE/CVF International Conference on Computer Vision, 2021, pp. 10 985–10 995

    2021

  21. [29]

    Mo- tiondiffuse: Text-driven human motion generation with diffusion model,

    M. Zhang, Z. Cai, L. Pan, F. Hong, X. Guo, L. Yang, and Z. Liu, “Mo- tiondiffuse: Text-driven human motion generation with diffusion model,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  22. [30]

    Temos: Generating diverse human motions from textual descriptions,

    M. Petrovich, M. J. Black, and G. Varol, “Temos: Generating diverse human motions from textual descriptions,” inEuropean Conference on Computer Vision. Springer, 2022, pp. 480–497

    2022

  23. [31]

    Gesturediffuclip: Gesture diffusion model with clip latents,

    T. Ao, Z. Zhang, and L. Liu, “Gesturediffuclip: Gesture diffusion model with clip latents,”ACM Transactions on Graphics, vol. 42, no. 4, pp. 1–18, 2023

    2023

  24. [32]

    Multi-semantics aggregation network based on the dynamic-attention mechanism for 3d human motion pre- diction,

    J. Shi, J. Zhong, and W. Cao, “Multi-semantics aggregation network based on the dynamic-attention mechanism for 3d human motion pre- diction,”IEEE Transactions on Multimedia, vol. 26, pp. 5194–5206, 2023

    2023

  25. [33]

    Character controllers using motion vaes,

    H. Y . Ling, F. Zinno, G. Cheng, and M. Van De Panne, “Character controllers using motion vaes,”ACM Transactions on Graphics (TOG), vol. 39, no. 4, pp. 40–1, 2020

    2020

  26. [34]

    A deep learning framework for character motion synthesis and editing,

    D. Holden, J. Saito, and T. Komura, “A deep learning framework for character motion synthesis and editing,”ACM Transactions on Graphics, vol. 35, no. 4, pp. 1–11, 2016

    2016

  27. [35]

    Physcap: Physically plausible monocular 3d motion capture in real time,

    S. Shimada, V . Golyanik, W. Xu, and C. Theobalt, “Physcap: Physically plausible monocular 3d motion capture in real time,”ACM Transactions on Graphics (ToG), vol. 39, no. 6, pp. 1–16, 2020

    2020

  28. [36]

    Neural monocular 3d human motion capture with physical awareness,

    S. Shimada, V . Golyanik, W. Xu, P. P ´erez, and C. Theobalt, “Neural monocular 3d human motion capture with physical awareness,”ACM Transactions on Graphics (ToG), vol. 40, no. 4, pp. 1–15, 2021

    2021

  29. [37]

    Simpoe: Simu- lated character control for 3d human pose estimation,

    Y . Yuan, S.-E. Wei, T. Simon, K. Kitani, and J. Saragih, “Simpoe: Simu- lated character control for 3d human pose estimation,” inProceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021, pp. 7159–7169

    2021

  30. [38]

    Physpt: Physics-aware pretrained transformer for estimating human dynamics from monocular videos,

    Y . Zhang, J. O. Kephart, Z. Cui, and Q. Ji, “Physpt: Physics-aware pretrained transformer for estimating human dynamics from monocular videos,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 2305–2317

    2024

  31. [39]

    Closely interactive human reconstruction with proxemics and physics-guided adaption,

    B. Huang, C. Li, C. Xu, L. Pan, Y . Wang, and G. H. Lee, “Closely interactive human reconstruction with proxemics and physics-guided adaption,” inProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 1011–1021

    2024

  32. [40]

    Human motion diffusion model,

    G. Tevet, S. Raab, B. Gordon, Y . Shafir, D. Cohen-or, and A. H. Bermano, “Human motion diffusion model,” inInternational Conference on Learning Representations, 2022

    2022

  33. [41]

    Smpl: A skinned multi-person linear model,

    M. Loper, N. Mahmood, J. Romero, G. Pons-Moll, and M. J. Black, “Smpl: A skinned multi-person linear model,”Acm Transactions on Graphics, vol. 34, no. Article 248, 2015

    2015

  34. [42]

    On the continuity of rotation representations in neural networks,

    Y . Zhou, C. Barnes, J. Lu, J. Yang, and H. Li, “On the continuity of rotation representations in neural networks,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2019, pp. 5745–5753

    2019

  35. [43]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” Advances in Neural Information Processing Systems, vol. 33, pp. 6840– 6851, 2020

    2020

  36. [44]

    Classifier-free diffusion guidance,

    J. Ho and T. Salimans, “Classifier-free diffusion guidance,” inNeurIPS Workshop on Deep Generative Models and Downstream Applications, 2021

    2021

  37. [45]

    Pvred: A position-velocity recurrent encoder-decoder for human motion prediction,

    H. Wang, J. Dong, B. Cheng, and J. Feng, “Pvred: A position-velocity recurrent encoder-decoder for human motion prediction,”IEEE Trans- actions on Image Processing, vol. 30, pp. 6096–6106, 2021

    2021

  38. [46]

    Proxycap: Real-time monocular full-body capture in world space via human-centric proxy-to-motion learning,

    Y . Zhang, H. Zhang, L. Hu, J. Zhang, H. Yi, S. Zhang, and Y . Liu, “Proxycap: Real-time monocular full-body capture in world space via human-centric proxy-to-motion learning,” inIEEE/CVF Conference on Computer Vision and Pattern Recognition, 2024, pp. 1954–1964

    2024

  39. [47]

    librosa: Audio and music signal analysis in python

    B. McFee, C. Raffel, D. Liang, D. P. Ellis, M. McVicar, E. Battenberg, and O. Nieto, “librosa: Audio and music signal analysis in python.” in SciPy, 2015, pp. 18–24

    2015

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.