Pith. sign in

REVIEW 3 major objections 4 minor 43 references

PINO: Person-Interaction Noise Optimization for Long-Duration and Customizable Motion Generation of Arbitrary-Sized Groups

T0 review · 3 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Pith's one-line read This paper proposes PINO, a training-free method that generates arbitrary-sized group motion by sequentially composing pairwise interactions from a pretrained two-person diffusion model, using noise-optimization penalties to keep bodies…

desk verdict A solid training-free composition method whose headline claim ('arbitrary-sized groups') outruns its evaluation, but the core idea and ablation are worth referee time. read the letter →

arxiv 2507.19292 v1 pith:6NY4HSLQ submitted 2025-07-25 cs.CV

classification cs.CV
keywords humanmotiongenerationmulti-personinteractiontext-conditioneddiffusionnoiseoptimizationinpaintingphysics-basedpenaltiestraining-freeinference
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

PINO tackles the problem of generating realistic motion for groups of arbitrarily many people without collecting multi-person training data. The central proposal is to break any group interaction into a sequence of pairwise interactions, each described by its own text prompt, and to build the group by adding one person at a time to a pretrained two-person interaction diffusion model. Because sequentially composed characters tend to overlap or penetrate each other, the method optimizes the diffusion model's initial noise against differentiable physics-based penalties that enforce minimum distances and user-specified spatial, orientational, and positional constraints. The same inpainting mechanism extends sequences in time, so long-duration motions with changing roles can be produced. Evaluations on up to five-person scenes show that the approach nearly eliminates overlap while keeping semantic fidelity comparable to the base model.

What carries the argument

The engine of the method is Person-Interaction Noise Optimization (PINO), defined as the minimization of differentiable penalty losses over the initial latent noise of a masked two-person diffusion model while one character's motion is fixed as the reference. The masked model $G^{\text{mask}}_\theta$ denoises only the new target character's noise, interleaving the reference character's already-generated frames at every step, so the new motion is anchored to the existing interaction. The loss $L = L_{\text{overlap}} + L_{\text{control}}$ penalizes frames in which two root positions come closer than a threshold $\delta$, plus user-specified control terms for root position, movement region, facing orientation, and relative distance, each written with thresholded max functions so that small deviations are free. For long-duration generation, motion inpainting blends a fixed reference window $m \odot \hat{x}$ with denoised continuation frames and adds a joint-acceleration boundary penalty for smooth transitions.

What would settle it

Run the method on a set of genuinely three-person-coordinated actions, such as a group hug, a two-person lift of a third, or a chain-reaction push, with prompts describing the full group behavior, and measure whether the resulting motion contains the required simultaneous three-body contact; if the composed pairwise motions systematically lose the third-person contact or produce physically impossible configurations, the pairwise-decomposition claim is falsified for those behaviors.

Watch

Extended reading notes

Core claim

On its own terms, the paper claims that group interactions are not monolithic: they are composites of smaller pairwise interactions linked by shared pivot characters, and therefore a model trained only for two-person interactions can generate coherent motion for groups of arbitrary size. The discovery is the demonstration that composing pairwise generations sequentially, with each new character conditioned on one existing character's frozen motion and guided by its own prompt, yields group motion whose quality does not collapse as people are added, provided the initial noise of the diffusion process is optimized with penalties for root-position overlap and, when needed, for target positions, movement regions, orientations, and relative distances. This noise optimization, not additional training, is what removes body penetration and overlap and what gives users fine-grained control. The paper further shows that the same framework extends motion in time via masked inpainting with a boundary acceleration penalty, allowing long sequences and prompt switching between pairs.

Load-bearing premise

The load-bearing premise is that every group interaction can be decomposed into a chain of independent pairwise interactions, each describable by its own text prompt; the paper's own limitations section concedes that coordinated three-body behaviors such as one person pushing another into a third are not captured by the two-person base model.

Editorial extensions

If this is right

  • Group size is no longer bounded by training data: the same frozen two-person base model can in principle be extended to any number of participants by repeating the pivot-and-optimize step.
  • Each pairwise interaction keeps its own text prompt, so mixed-role scenes such as a photographer directing posers or dancers with onlookers can be specified directly instead of being forced into one shared group description.
  • New control signals, including where someone stands, where they face, how fast they move, and how far apart they are, become inference-time penalty choices, so no per-control retraining is needed.
  • Long sequences with changing roles over time are reachable by extending motions through inpainting while switching prompts between pairs, as demonstrated by the alternating handshake among three people.
  • The overlap and penetration artifacts that grow with group size in sequential baselines are largely suppressed while FID, diversity, and R-precision stay close to the underlying two-person model.

Reading between the lines

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

  • Because the penalties act only on joint and root positions and use a single distance threshold, the method implicitly treats bodies as point-like at the root; a natural extension would couple body-thickness buffers or hand-aware losses to catch the residual near-hand penetrations the paper reports.
  • The same training-free recipe could turn any two-person motion prior into a data-generation tool: run PINO over diverse prompts and pairings, then use the artifact-free outputs to train a feed-forward multi-person model, which would remove the per-person optimization cost at deployment.
  • The pairwise-decomposition assumption means group behaviors whose dynamics are inherently third-order, such as a lift, a chain-reaction fall, or a three-person synchronized jump, will be under-expressed; testing exactly where the dyadic graph fails would map the boundary of the method.
  • The control penalties are generic differentiable costs, so in principle the same optimizer could steer motions to satisfy non-human constraints such as camera framing, obstacle maps, or collision with scene geometry, not just person-person spacing.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 4 minor

Summary. The paper proposes PINO, a training-free framework that extends a pretrained two-person motion-diffusion model (InterGen) to multi-person group generation by sequentially composing pairwise interactions. Each new character is generated while conditioning on a fixed reference character, and inference-time noise optimization applies penalty terms for overlap avoidance and user-controllable spatial/temporal constraints (root position, region, orientation, relative distance). Motion inpainting is added to extend sequences in time, and prompt switching between pairs enables changing roles during long sequences. Experiments report lower overlap and penetration volume on two-person, multi-person, and long-duration tasks, an ablation of the penalty terms, a small user study, and comparisons with InterGen, FreeMotion, and in2IN.

Significance. If the claims hold, PINO is a practically useful, training-free way to obtain customizable multi-person motion without dedicated multi-person datasets, which is valuable for animation and content generation. The paper's strengths include using a publicly available pretrained model, providing fairly detailed supplementary implementation details, reporting a user study, and demonstrating that the proposed framework also works with another two-person base model (in2IN). The significance is tempered, however, by two issues: the headline overlap metric is a strict relaxation of the optimized penalty, so the main quantitative success is partly by construction, and the semantic evaluation of multi-person output decomposes everything into pairs sharing a single pivot, which cannot validate the paper's central claim about coherent group interactions of arbitrary size.

major comments (3)
  1. [Sec. 4.2, Eqs. (6)-(7); Sec. 5.2, Table 4; Sec. 6] The central claim of generating group interactions of arbitrary size is not validated at the group level. Equations (6)-(7) generate each new character conditioned on the fixed reference motion xhat^{k_p}_0, and that reference is never updated when the new character is added. As Section 6 concedes, a person pushing another person into a third person is therefore outside the model. The semantic evaluation in Table 4 only decomposes each multi-person sequence into the pairs (1,2), ..., (1,5), all sharing person 1 as pivot, and computes FID against two-person ground truth; this protocol cannot detect whether the generated characters perform a coherent group-level action, and for FreeMotion it likely understates that baseline's group-level conditioning. Please add a group-level semantic evaluation (for example, recognition of multi-person actions, evaluation on non-pivot pairings, or explicit tests of synchronized group behaviors such as group lifts or cascading falls) and temper the abstract and conclusion claims accordingly.
  2. [Sec. 4.3, Eq. (8); Supplementary B.1-B.2; Tables 2-4] The overlap numbers that form the main quantitative evidence are by construction determined by the optimization objective. Loverlap uses a root-distance threshold delta_overlap=30cm (Supplementary B.1), while the Overlap metric declares a failure when root positions are closer than 25cm (Supplementary B.2). The penalty is thus a stricter version of the metric, so Overlap values near zero in Tables 2-4 certify that the optimizer minimized its own loss, not that the framework produces physically plausible interactions in an independent sense. The remaining non-semantic metrics are more informative; note that PINO's Foot Skate and Max Acc in Table 4 are consistently worse at (1,4) and (1,5) than the InterGen baseline, a degradation that receives no discussion. Reporting a metric that is a strict relaxation of the objective, and discussing the metrics that do move unfavorably, would put the physical-plausibility claim on firmer ground.
  3. [Sec. 5.2, Tables 3-4; Supplementary B.1 and C.4] The comparison with InterGen and FreeMotion is not matched for inference-time optimization: PINO performs up to 100 gradient steps of noise optimization per added character, while the baselines are run without the additional optimization loop, with PINO taking about one minute per person for the overlap penalty alone (Supplementary C.4). The large overlap reduction may therefore reflect the optimization budget rather than the pairwise-composition design. A fairer assessment would include a compute-matched variant (for example, optimizing the baselines' latent initialization with the same loss, or an ablation of PINO without the overlap penalty) so the reader can separate the contribution of sequential composition from that of noise optimization.
minor comments (4)
  1. [Sec. 4.2 and Algorithm 1] The notation for the number of individuals is inconsistent: Section 4.2 writes x3,...,xN while Algorithm 1 uses M, and N is already used for the number of frames in the motion representation. Please unify these symbols.
  2. [Algorithm 1, line 12 and Eq. (7)] Line 12 of Algorithm 1 calls Gθ(...) whereas Eq. (7) uses Gmask_theta, and the argument order in the pseudocode appears different from the convention used in Eqs. (3)-(7). Please make the pseudocode consistent with the equations.
  3. [Figure B (Supplementary C.3)] The bar labels in the user-study figure run together as "InterGenFreeMotionPINO", and the reported scores are not accompanied by error bars or significance tests across the 35 participants. Please report confidence intervals or a statistical comparison.
  4. [Supplementary B.1-B.2] The overlap metric threshold of 25 cm and the penalty threshold of 30 cm appear only in the supplementary material; since these values directly determine the headline results, they should be stated in the main text.

Circularity Check

2 steps flagged · score 4.0 of 10

Overlap reduction is built into the optimized loss, and the group-semantic metric is defined by the same pairwise decomposition the method assumes.

  1. fitted input called prediction [Section 4.3, Eq. 8; Section 5.2 Table 2 and Tables 3/4; Supplementary B.1, B.2]
    "we introduce Loverlap that penalizes the root positions of individuals if they are closer than a predefined threshold δ (Sec. 4.3); 'we define an overlap occurrence as any frame in the motion sequence where the root positions of any pair of individuals are closer than 25 cm' (Sec. B.2); 'The overlap penalty threshold, δoverlap, is set to 30 cm' (Sec. B.1)."

    Minimizing Eq. 8 directly drives the paper's Overlap metric: the penalty is active whenever two root positions are closer than 30 cm, while the reported metric counts overlaps at 25 cm. A successfully optimized Loverlap therefore produces Overlap≈0 by construction. Reporting this value in Tables 2-4 as evidence of physical plausibility is a check that the loss was minimized, not an independent prediction. Penetration Volume, FID, Foot Skate, and Max Acc. are not directly optimized, which limits the scope of the circularity.

  2. self definitional [Section 4.2 Eqs. 6-7; Section 5.2 Table 4; Section 6 Limitations]
    "we evaluate semantic correctness by decomposing each multi-person sequence into two-person pairs and computing FID, overlap, foot skating, and Max Acc. for every pair (Sec. 5.2); 'its focus on pairwise interactions may not fully capture the dynamics of coordinated group behaviors. For example, a person pushing another person into a third person will not be modeled by the underlying two-person model' (Sec. 6)."

    Eqs. 6-7 generate each new person p only against a fixed reference x̂^{k_p}_0, never updating the reference in response to the newcomer. The only group-level semantic evaluation then scores exactly the pairs (1,p) formed with the first character as pivot. A coordinated group behavior that requires the pivot to react physically (e.g., A pushes B into C) is unrepresentable by Eqs. 6-7, and the pair-FID cannot detect its failure. Thus the semantic 'multi-person' evaluation is defined in terms of the same pairwise decomposition used to construct the output, so it cannot validate arbitrary-size group coordination.

full rationale

PINO is a training-free inference-time method: it composes a pretrained two-person diffusion model (InterGen) sequentially, optimizing initial noise with penalties. The main circular components are two. First, the headline overlap improvement is the direct target of the Loverlap loss, so near-zero overlap scores in Tables 2-4 are expected by construction rather than independent evidence. Second, the semantic evaluation of multi-person output (Table 4) decomposes every generated sequence into pairs with a fixed pivot, which is exactly the generative unit used in Eqs. 6-7; this metric cannot distinguish a coherent group interaction from a set of pairwise interactions, and the paper's own Limitations section concedes that coordinated behaviors such as a person pushing another into a third are outside the model. These are genuine partial circularities. However, the paper also reports metrics not optimized by Loverlap—Foot Skate, Max Acc., FID, Diversity, R-Precision, and a user study—which provide partially independent support for motion quality and text faithfulness. There is no load-bearing self-citation: the base model InterGen and the noise-optimization framework ProgMoGen come from other research groups, and the author's own RIG [30] is only mentioned in related work. The central claim is therefore not forced by a self-citation chain or by definition, but a portion of the quantitative evidence reduces to the optimized objective and the pairwise evaluation assumption.

Assumptions & free parameters 5 free parameters · 3 assumptions · 0 invented entities

The central claim rests on the pretrained two-person diffusion model (InterGen) behaving as a valid prior under a masked, reference-conditioned modification, and on the assumption that group interactions decompose into pairwise interactions. Hand-chosen thresholds and optimization schedules are free parameters that affect the results but are not derived.

free parameters (5)
  • Overlap penalty distance threshold = 30 cm
    Hand-chosen in Section B.1 to enforce spatial separation; the overlap evaluation threshold is 25 cm, so the penalty threshold is looser than the metric.
  • Orientation penalty threshold = 0.2
    Cosine-similarity margin in Eq. 14, set by hand in Section B.1.
  • Learning rate for noise optimization = 0.003
    Adam learning rate for noise optimization (Section 5.1), chosen by trial; no sensitivity analysis is provided.
  • Number of optimization steps = 100 (early stop at loss < 1e-6)
    Budget for the inner loop; the early stopping threshold is also ad hoc (Sections 5.1 and B.1).
  • Relative weights of penalty terms = not reported
    The paper defines L = L_overlap + L_control (Eq. 9) without listing coefficients for combining multiple control terms; this is a missing detail for reproduction.
assumptions (3)
  • domain assumption A pretrained two-person interaction diffusion model (InterGen) can serve as a general prior for arbitrary group compositions, including the masked reference-conditioned generation G_mask_theta.
    PINO conditions the generation of each new character on a fixed reference character's motion by injecting a noisy version of the reference at every denoising step (Section 4.2, Eq. 6). InterGen was trained to generate both characters jointly, so this single-character conditioning is a modification whose validity is assumed rather than demonstrated.
  • domain assumption Any group interaction can be decomposed into a sequence of pairwise interactions anchored on shared pivot characters.
    This is the key structural assumption of the method (Sections 4.1 and 4.2) and is explicitly acknowledged as a limitation in Section 6: higher-order coordinated behaviors such as pushing one person into a third are not captured.
  • domain assumption Optimizing the initial diffusion noise with differentiable penalties preserves text-prompt alignment and natural motion quality.
    The paper assumes that 100 steps of Adam on the latent noise will reduce penalties without breaking the prompt-motion correspondence; empirical FID and R-precision are roughly unchanged for two-person generation, but this is not guaranteed across prompts (Section 5.2, Table 2).

how reviews work

0 comments
Cite this review

Pith. "Pith review of PINO: Person-Interaction Noise Optimization for Long-Duration and Customizable Motion Generation of Arbitrary-Sized Groups." pith.science (2026). https://pith.science/paper/6NY4HSLQ

@misc{pith2026250719292,
  author       = {Pith},
  title        = {Pith review of: PINO: Person-Interaction Noise Optimization for Long-Duration and Customizable Motion Generation of Arbitrary-Sized Groups},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/6NY4HSLQ}},
  note         = {Machine review of arXiv:2507.19292}
}
read the original abstract

Generating realistic group interactions involving multiple characters remains challenging due to increasing complexity as group size expands. While existing conditional diffusion models incrementally generate motions by conditioning on previously generated characters, they rely on single shared prompts, limiting nuanced control and leading to overly simplified interactions. In this paper, we introduce Person-Interaction Noise Optimization (PINO), a novel, training-free framework designed for generating realistic and customizable interactions among groups of arbitrary size. PINO decomposes complex group interactions into semantically relevant pairwise interactions, and leverages pretrained two-person interaction diffusion models to incrementally compose group interactions. To ensure physical plausibility and avoid common artifacts such as overlapping or penetration between characters, PINO employs physics-based penalties during noise optimization. This approach allows precise user control over character orientation, speed, and spatial relationships without additional training. Comprehensive evaluations demonstrate that PINO generates visually realistic, physically coherent, and adaptable multi-person interactions suitable for diverse animation, gaming, and robotics applications.

Figures

Figures reproduced from arXiv: 2507.19292 by the authors.

Figure 1
Figure 1. Person-Interaction Noise Optimization (PINO) leverages pre-trained two-person interaction diffusion models as motion priors to generate motions of arbitrary-sized groups with the combination of spatiotemporal penalty, prompt switching between pairs, and motion extension. The first example employs penalties to guide humans into the form “CV”, the second generates a pair posing for a photograph, then switches the prom… view at source ↗
Figure 2
Figure 2. Person-Interaction Noise Optimization (PINO). Red marks the already-generated reference motions, whereas purple high￾lights the elements for the newly added person. A masked diffusion model transforms the latent noise of a new individual, which is guided by a text prompt and reference motion, into an initial motion that is subsequently refined through noise optimization. This process is then repeated for groups of a… view at source ↗
Figure 3
Figure 3. Comparison of the generated multi-person interaction. [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Qualitative results of multi-person interaction genera [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Visualization of the ablation study. We incrementally [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 33 canonical work pages

  1. [1]

    Teach: Temporal action composition for 3d hu- mans

    Nikos Athanasiou, Mathis Petrovich, Michael J Black, and G¨ul Varol. Teach: Temporal action composition for 3d hu- mans. In 3DV, 2022. 2

  2. [2]

    Seamless human motion composition with blended posi- tional encodings

    German Barquero, Sergio Escalera, and Cristina Palmero. Seamless human motion composition with blended posi- tional encodings. In CVPR, 2024. 2, 3

  3. [3]

    Attend-and-excite: Attention-based se- mantic guidance for text-to-image diffusion models

    Hila Chefer, Yuval Alaluf, Yael Vinker, Lior Wolf, and Daniel Cohen-Or. Attend-and-excite: Attention-based se- mantic guidance for text-to-image diffusion models. ACM TOG, 42(4):1–10, 2023. 2, 3

  4. [4]

    Executing your commands via motion diffusion in latent space

    Xin Chen, Biao Jiang, Wen Liu, Zilong Huang, Bin Fu, Tao Chen, and Gang Yu. Executing your commands via motion diffusion in latent space. In CVPR, 2023. 2

  5. [5]

    Motionlcm: Real-time controllable motion generation via latent consistency model

    Wenxun Dai, Ling-Hao Chen, Jingbo Wang, Jinpeng Liu, Bo Dai, and Yansong Tang. Motionlcm: Real-time controllable motion generation via latent consistency model. In ECCV,

  6. [6]

    Freemotion: A unified framework for number- free text-to-motion synthesis

    Ke Fan, Junshu Tang, Weijian Cao, Ran Yi, Moran Li, Jingyu Gong, Jiangning Zhang, Yabiao Wang, Chengjie Wang, and Lizhuang Ma. Freemotion: A unified framework for number- free text-to-motion synthesis. In ECCV, 2024. 2, 3, 7

  7. [7]

    Synthesis of compositional animations from textual descriptions

    Anindita Ghosh, Noshaba Cheema, Cennet Oguz, Christian Theobalt, and Philipp Slusallek. Synthesis of compositional animations from textual descriptions. In ICCV, 2021. 2

  8. [8]

    Generating diverse and natural 3d human motions from text

    Chuan Guo, Shihao Zou, Xinxin Zuo, Sen Wang, Wei Ji, Xingyu Li, and Li Cheng. Generating diverse and natural 3d human motions from text. In CVPR, 2022. 2, 3

Show all 43 references
  1. [9]

    Momask: Generative masked mod- eling of 3d human motions

    Chuan Guo, Yuxuan Mu, Muhammad Gohar Javed, Sen Wang, and Li Cheng. Momask: Generative masked mod- eling of 3d human motions. In CVPR, 2024. 2

  2. [10]

    Initno: Boosting text-to-image diffu- sion models via initial noise optimization

    Xiefan Guo, Jinlin Liu, Miaomiao Cui, Jiankai Li, Hongyu Yang, and Di Huang. Initno: Boosting text-to-image diffu- sion models via initial noise optimization. In CVPR, 2024. 2, 3

  3. [11]

    Denoising dif- fusion probabilistic models

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. Denoising dif- fusion probabilistic models. In NeurIPS, pages 6840–6851,

  4. [12]

    Learned motion matching

    Daniel Holden, Oussama Kanoun, Maksym Perepichka, and Tiberiu Popa. Learned motion matching. ACM TOG, 39(4): 53–1, 2020. 2

  5. [13]

    Intermask: 3d human interaction genera- tion via collaborative masked modelling

    Muhammad Gohar Javed, Chuan Guo, Li Cheng, and Xingyu Li. Intermask: 3d human interaction genera- tion via collaborative masked modelling. arXiv preprint arXiv:2410.10010, 2024. 3

  6. [14]

    Action-gpt: Leveraging large-scale language models for improved and generalized zero shot ac- tion generation

    Sai Shashank Kalakonda, Shubh Maheshwari, and Ravi Ki- ran Sarvadevabhatla. Action-gpt: Leveraging large-scale language models for improved and generalized zero shot ac- tion generation. arXiv preprint arXiv:2211.15603, 2022. 2

  7. [15]

    Guided motion diffusion for controllable human motion synthesis

    Korrawe Karunratanakul, Konpat Preechakul, Supasorn Suwajanakorn, and Siyu Tang. Guided motion diffusion for controllable human motion synthesis. In ICCV, pages 2151– 2162, 2023. 2

  8. [16]

    Op- timizing diffusion noise can serve as universal motion priors

    Korrawe Karunratanakul, Konpat Preechakul, Emre Aksan, Thabo Beeler, Supasorn Suwajanakorn, and Siyu Tang. Op- timizing diffusion noise can serve as universal motion priors. In CVPR, 2024. 2, 3

  9. [17]

    Interhandgen: Two-hand interaction generation via cascaded reverse diffusion

    Jihyun Lee, Shunsuke Saito, Giljoo Nam, Minhyuk Sung, and Tae-Kyun Kim. Interhandgen: Two-hand interaction generation via cascaded reverse diffusion. In CVPR, 2024. 6

  10. [18]

    T2lm: Long-term 3d human mo- tion generation from multiple sentences

    Taeryung Lee, Fabien Baradel, Thomas Lucas, Kyoung Mu Lee, and Gr `egory Rogez. T2lm: Long-term 3d human mo- tion generation from multiple sentences. In CVPR, 2024. 3

  11. [19]

    Intergen: Diffusion-based multi-human motion genera- tion under complex interactions

    Han Liang, Wenqian Zhang, Wenxuan Li, Jingyi Yu, and Lan Xu. Intergen: Diffusion-based multi-human motion genera- tion under complex interactions. IJCV, pages 1–21, 2024. 2, 3, 4, 6, 13

  12. [20]

    Programmable motion generation for open- set motion control tasks

    Hanchao Liu, Xiaohang Zhan, Shaoli Huang, Tai-Jiang Mu, and Ying Shan. Programmable motion generation for open- set motion control tasks. In CVPR, 2024. 3, 11

  13. [21]

    Smpl: A skinned multi- person linear model

    Matthew Loper, Naureen Mahmood, Javier Romero, Gerard Pons-Moll, and Michael J Black. Smpl: A skinned multi- person linear model. ACM TOG, 2015. 6

  14. [22]

    Black, and G ¨ul Varol

    Mathis Petrovich, Michael J. Black, and G ¨ul Varol. Action- conditioned 3d human motion synthesis with transformer vae. In ICCV, 2021. 2

  15. [23]

    Temos: Generating diverse human motions from textual descriptions

    Mathis Petrovich, Michael J Black, and G ¨ul Varol. Temos: Generating diverse human motions from textual descriptions. In ECCV, 2022. 2

  16. [24]

    Mmm: Generative masked motion model

    Ekkasit Pinyoanuntapong, Pu Wang, Minwoo Lee, and Chen Chen. Mmm: Generative masked motion model. In CVPR,

  17. [25]

    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, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In ICML, 2021. 2

  18. [26]

    in2in: Leveraging individual information to generate human interactions

    Pablo Ruiz-Ponce, German Barquero, Cristina Palmero, Ser- gio Escalera, and Jos´e Garc´ıa-Rodr´ıguez. in2in: Leveraging individual information to generate human interactions. In CVPRW, 2024. 3, 13

  19. [27]

    Human motion diffusion as a generative prior

    Yoni Shafir, Guy Tevet, Roy Kapon, and Amit Haim Bermano. Human motion diffusion as a generative prior. In ICLR, 2024. 2, 3

  20. [28]

    Towards open domain text-driven synthesis of multi-person motions

    Mengyi Shan, Lu Dong, Yutao Han, Yuan Yao, Tao Liu, Ifeoma Nwogu, Guo-Jun Qi, and Mitch Hill. Towards open domain text-driven synthesis of multi-person motions. In ECCV, 2024. 2, 3

  21. [29]

    Denois- ing diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denois- ing diffusion implicit models. In ICLR, 2021. 4

  22. [30]

    Role-aware interaction generation from textual description

    Mikihiro Tanaka and Kent Fujiwara. Role-aware interaction generation from textual description. In ICCV, 2023. 2, 3

  23. [31]

    Human motion diffu- sion model

    Guy Tevet, Sigal Raab, Brian Gordon, Yoni Shafir, Daniel Cohen-Or, and Amit Haim Bermano. Human motion diffu- sion model. In ICLR, 2023. 2, 3 9

  24. [32]

    Karen Liu

    Jonathan Tseng, Rodrigo Castellon, and C. Karen Liu. Edge: Editable dance generation from music. In CVPR, 2023. 3

  25. [33]

    Intercontrol: Zero-shot human interaction generation by controlling every joint

    Zhenzhi Wang, Jingbo Wang, Yixuan Li, Dahua Lin, and Bo Dai. Intercontrol: Zero-shot human interaction generation by controlling every joint. In NeurIPS, 2024. 2, 3

  26. [34]

    Omnicontrol: Control any joint at any time for human motion generation

    Yiming Xie, Varun Jampani, Lei Zhong, Deqing Sun, and Huaizu Jiang. Omnicontrol: Control any joint at any time for human motion generation. In ICLR, 2024. 2

  27. [35]

    Actformer: A gan-based transformer towards general action-conditioned 3d human motion gener- ation

    Liang Xu, Ziyang Song, Dongliang Wang, Jing Su, Zhicheng Fang, Chenjing Ding, Weihao Gan, Yichao Yan, Xin Jin, Xi- aokang Yang, et al. Actformer: A gan-based transformer towards general action-conditioned 3d human motion gener- ation. In ICCV, 2023. 2

  28. [36]

    Inter-x: Towards versatile human- human interaction analysis

    Liang Xu, Xintao Lv, Yichao Yan, Xin Jin, Shuwen Wu, Congsheng Xu, Yifan Liu, Yizhou Zhou, Fengyun Rao, Xingdong Sheng, et al. Inter-x: Towards versatile human- human interaction analysis. In CVPR, 2024. 3

  29. [37]

    Convolutional sequence generation for skeleton- based action synthesis

    Sijie Yan, Zhizhong Li, Yuanjun Xiong, Huahan Yan, and Dahua Lin. Convolutional sequence generation for skeleton- based action synthesis. In CVPR, 2019. 2

  30. [38]

    T2m-gpt: Generating human motion from textual de- scriptions with discrete representations

    Jianrong Zhang, Yangsong Zhang, Xiaodong Cun, Shaoli Huang, Yong Zhang, Hongwei Zhao, Hongtao Lu, and Xi Shen. T2m-gpt: Generating human motion from textual de- scriptions with discrete representations. In CVPR, 2023. 2

  31. [39]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. In ICCV, 2023. 2, 3

  32. [40]

    Re- modiffuse: Retrieval-augmented motion diffusion model

    Mingyuan Zhang, Xinying Guo, Liang Pan, Zhongang Cai, Fangzhou Hong, Huirong Li, Lei Yang, and Ziwei Liu. Re- modiffuse: Retrieval-augmented motion diffusion model. In ICCV, 2023. 2

  33. [41]

    Motiondif- fuse: Text-driven human motion generation with diffusion model

    Mingyuan Zhang, Zhongang Cai, Liang Pan, Fangzhou Hong, Xinying Guo, Lei Yang, and Ziwei Liu. Motiondif- fuse: Text-driven human motion generation with diffusion model. IEEE TPAMI, 46(6):4115–4128, 2024. 2

  34. [42]

    Diffcollage: Parallel generation of large content with diffusion models

    Qinsheng Zhang, Jiaming Song, Xun Huang, Yongxin Chen, and Ming yu Liu. Diffcollage: Parallel generation of large content with diffusion models. In CVPR, 2023. 3

  35. [43]

    the other person approaches one by walking

    Rui Zhao, Hui Su, and Qiang Ji. Bayesian adversarial human motion synthesis. In CVPR, 2020. 2 10 PINO: Person-Interaction Noise Optimization for Long-Duration and Customizable Motion Generation of Arbitrary-Sized Groups Supplementary Material A. Formulation of Penalty Function...

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.