Pith. sign in

REVIEW 5 major objections 5 minor 18 references

Co$^{3}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Two-speaker co-speech gestures can be generated coherently from conversational audio, and Co3Gesture backs this with a 7M-frame whole-body dataset.

desk verdict The dataset is a real contribution; the model's reported wins are plausible but rest on a self-contained pseudo-label benchmark, and the user-study statistics look inconsistent. read the letter →

arxiv 2505.01746 v1 pith:C5LQ7RQA submitted 2025-05-03 cs.CV

Xingqun Qi , Yatian Wang , Hengyuan Zhang , Jiahao Pan , Wei Xue , Shanghang Zhang , Wenhan Luo , Qifeng Liu
show 1 more author
Yike Guo
This is my paper · ORCID
classification cs.CV
keywords co-speechgesturegenerationconcurrenttwo-speakergesturesdiffusionmodelstemporalinteractionmoduleGES-Interdatasetwhole-body3Dmeshposespeakerdiarizationmutualattention
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

Co3Gesture takes on a task that previous co-speech gesture models skip: generating the gestures of both people at once during a two-person conversation, rather than animating one speaker in isolation. To support this, the paper builds GES-Inter, a dataset of more than seven million frames of two-person whole-body mesh poses with separated audio, transcripts, and phoneme timestamps extracted from talk shows and interviews. The method uses two cooperative diffusion branches, one per speaker, each conditioned on that speaker's separated audio, plus a Temporal Interaction Module that blends mixed-audio interaction cues into each branch and a mutual-attention layer that lets the two gesture streams inform each other. On GES-Inter, the paper reports lower Fréchet Gesture Distance and higher beat consistency and diversity than single-speaker co-speech and text-to-motion baselines. If those results hold, avatar animation could move from monologue gestures to realistic dyadic conversation.

What carries the argument

The load-bearing mechanism is the Temporal Interaction Module (TIM), a cross-attention unit that, for each speaker, takes the speaker's own audio embedding as a query against their motion features, then builds a temporal correlation matrix between the self-conditioned motion and a motion embedding conditioned on the mixed conversation audio. A sigmoid-gated learned weight balances the two, so each branch's gesture features are softly steered by the interaction context. On top of this, a mutual attention layer exchanges query information between the two branches, and shared weight updates across the bilateral denoisers exploit the exchange symmetry of the two speakers. The full system is trained with a diffusion denoising objective plus velocity and foot-contact losses.

What would settle it

Collect a small validation set of true motion-capture recordings of two-person conversations, or manually cleaned pose tracks from the same videos, and compare Co3Gesture's outputs against them with the same FGD and beat-consistency metrics; if the performance gap over baselines shrinks or vanishes on clean ground truth, the pseudo-label pipeline was doing the work. Alternatively, measure motion amplitude and wrist-angle distributions in GES-Inter against an unfiltered sample to test whether the filtering selected unusually static clips.

Watch

Extended reading notes

Core claim

The paper's central claim is that coherent concurrent two-speaker gestures can be generated directly from conversational audio, and that the right way to do this is to keep the two speakers' generation separate but coupled. It constructs GES-Inter, a large-scale pseudo-labelled dataset of two-person interactive whole-body poses, and proposes Co3Gesture: two transformer-based diffusion denoisers, each generating one speaker's upper-body motion from that speaker's separated voice, while a Temporal Interaction Module fuses features from the mixed conversation audio into each branch as an interaction guide. A mutual attention mechanism then lets the two denoising streams exchange information. The paper reports that this design outperforms all compared single-speaker co-speech and text-to-motion baselines on GES-Inter, and that ablations removing the interaction module, mutual attention, bilateral branches, or mixed-audio conditioning all degrade the metrics.

Load-bearing premise

The whole benchmark rests on the assumption that the automatically extracted and heavily filtered 3D poses from talk-show videos are accurate and representative enough to serve as ground truth for both training and evaluation; if those pseudo-labels contain systematic noise or the filtering kept only unusually static clips, the reported gains may not transfer to real conversational motion.

Editorial extensions

If this is right

  • Two-speaker conversational gesture synthesis becomes a tractable generation task from audio alone, so virtual avatars can gesture in dialogue rather than only in monologue.
  • GES-Inter gives the research community a large-scale whole-body mesh dataset with separated audio, transcripts, and phoneme timestamps, enabling downstream work on talking-face generation, behavior analysis, and dyadic motion modeling.
  • The bilateral-branch design with a temporal interaction module is claimed to be necessary: replacing TIM with an MLP, removing mutual attention, or collapsing to a single holistic branch all raise FGD and lower beat consistency.
  • Using both separated speaker audio and mixed conversation audio as conditions is claimed to preserve speech-rhythm alignment while keeping the two gesture streams interactively coherent.

Reading between the lines

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

  • If the pseudo-label noise is systematic, the benchmark may reward models that reproduce smoothed, filtered motion rather than genuine conversational behavior; a small motion-capture validation split would settle this.
  • The soft-weight fusion idea in TIM could generalize to group conversations by chaining pairwise interaction modules across more than two speakers.
  • The same two-branch diffusion plus interaction-guidance recipe is a plausible template for other dyadic motion tasks, such as two-person dance or turn-taking body language, by swapping the audio condition for music or another cue.
  • Reporting the retention rate of the filtering pipeline, that is, how many raw clips survive each processing step, would let future work judge how much of GES-Inter's quality comes from selection rather than modeling.
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

5 major / 5 minor

Summary. The paper introduces GES-Inter, a large-scale pseudo-labeled dataset of two-speaker concurrent co-speech gestures extracted from in-the-wild talk shows and interviews, containing more than 7M frames and 70 hours of SMPL-X whole-body meshes with separated per-speaker audio, transcripts, and phoneme alignments. On top of this dataset, the authors propose Co3Gesture, a bilateral diffusion framework with two cooperative branches conditioned on the separated speaker audios, a Temporal Interaction Module (TIM) that fuses mixed-audio interaction cues into each branch, and a mutual attention mechanism intended to improve inter-speaker coherence. Experiments on GES-Inter report FGD, BC, Diversity, ablations, and a user study, claiming state-of-the-art performance over single-person co-speech and text2motion baselines. The paper also explicitly acknowledges that pose-extraction noise and upper-body-only generation are limitations.

Significance. If the claims hold, the paper makes two useful contributions: GES-Inter is, to my knowledge, the first large-scale mesh-based whole-body concurrent two-speaker co-speech gesture dataset, and the bilateral diffusion design with temporal interaction fusion is a plausible architecture for asymmetric dyadic gesture generation. The dataset and code are promised to be public, which is a concrete asset for the community. However, the significance is currently conditional: all quantitative evidence is obtained on the authors' own dataset, the mutual-attention mechanism is not formally specified, and the adaption of single-person baselines to the two-person setting is not described. These gaps prevent the reader from verifying the central superiority claim, so the contributions are not yet fully established.

major comments (5)
  1. [Sec. 4.2 (SOTA comparisons)] The paragraph on SOTA comparisons states that all competitors are implemented from official source codes or pretrained models and that the text2motion counterparts receive 'the same audio encoder as ours,' but it does not specify how the single-person co-speech models (TalkSHOW, ProbTalk, DiffSHEG, EMAGE) are adapted to produce two simultaneous speakers, nor how the text-driven models (MDM, InterX, InterGen) are conditioned on audio rather than text or action labels. Without these adaptation details, Table 2 is not reproducible and the claim of a fair comparison is unsupported. Please give the exact input/output configuration for each baseline (e.g., per-speaker separated audio passed independently through a single-speaker model, two-branch variants, and where the audio encoder is inserted).
  2. [Sec. 3.3 (Mutual Attention Mechanism)] The mutual attention mechanism is described only in words: each branch is said to take the counterpart's features as the query in multi-head attention, with shared weight update strategies, but no equations, tensor shapes, or combination rules are provided. Since mutual attention is one of the two core technical contributions (abstract, contributions, and ablation Table 3), the method is not reproducible as written. Please provide the formal definition of the mutual attention operation, including how Q, K, V are formed from the two branch features and how the attended output is merged with the existing branch embeddings.
  3. [Sec. 3.4 and Table 5 (foot contact loss)] The paper defines L_foot following Tevet et al. and Table 5 reports that removing it degrades FGD from 0.769 to 1.082 and BC from 0.692 to 0.675. However, Section 4.1 and Appendix A.2 state that only upper-body joints are generated, and the ablation note for Table 5 says the lower body is completed as a T-pose in the forward-kinematic function used to compute the loss. With a fixed T-pose lower body, it is unclear what physical quantity the foot contact loss regularizes or how it can affect the generated upper-body motion. Please clarify the exact loss definition and explain the mechanism by which this loss produces such a large improvement.
  4. [Sec. 4 (evaluation scope)] All quantitative comparisons are performed on GES-Inter, the authors' own dataset, and the FGD feature extractor is presumably trained on GES-Inter as well. Because the pseudo-label pipeline (Appendix A.1.1) aggressively filters out fast, extreme, or jittery motions (wrist Euler angles above 150 degrees rejected, inter-frame changes above 25 degrees at 15 fps rejected, plus manual group-level discarding), the measured distribution may not be representative of conversational gestures in other settings. To support the claim of general superiority, please evaluate at least one existing dyadic or conversational benchmark (e.g., TWH16.2, BEAT2, or the GENEA 2023 dyadic setting) and describe how the FGD autoencoder is trained.
  5. [Tables 2-5 (confidence intervals)] The tables report 95% confidence intervals only for Diversity; FGD and BC are given as point estimates. FGD is stochastic and known to be sensitive to the evaluation sample, and some reported gaps (e.g., 0.769 vs 1.012 in Table 2; 0.924 vs 0.769 in Table 3) may be within noise. Please report confidence intervals or standard deviations over multiple evaluation seeds or bootstrap resamples for FGD and BC, and state the number of resamples used.
minor comments (5)
  1. [Eq. (3)] Equation (3) appears to be missing norm notation; the terms should be written as ||x_a - D(x_a^(t), C_a, C_mix, t)||_2^2 and similarly for speaker b.
  2. [Sec. 3.4] The word 'hyper-parameter' is misspelled as 'hper-parameter' in the sentence defining sigma(t).
  3. [Sec. 3.1] The citation for the MANO hand model is attributed to Boukhayma et al. (2019); the canonical MANO reference is Romero et al. (2017), and the current citation should be corrected.
  4. [Appendix A.3] The user study section would benefit from a statement about whether the t-test was corrected for multiple comparisons across the seven paired comparisons (one per competitor).
  5. [Conclusion] The explicit limitation statement regarding pose-extraction noise and upper-body-only generation is appreciated; in the journal version, please connect this limitation to the external-validation request in the evaluation section.

Circularity Check

0 steps flagged · score 0.0 of 10

No derivation-level circularity: the diffusion losses, audio-conditioned branches, and interaction modules are defined independently of the reported metrics.

full rationale

The paper's derivation chain is self-contained rather than circular. Given separated speaker audio Ca, Cb and mixed audio Cmix, the bilateral diffusion denoisers predict xa and xb with the standard objective Lsimple = E || x - D(x(t), C, Cmix, t) ||^2, plus velocity and foot-contact losses; none of these losses or architectural components is defined in terms of FGD, BC, or Diversity, so the numerical results are not forced by construction. The Temporal Interaction Module and mutual attention mechanism are learned modules based on cross-attention and a sigmoid-gated fusion, not fitted to the benchmark numbers. The GES-Inter dataset is constructed with external tools (PyMAF-X, pyannote-audio, WhisperX, MFA, SmoothNet) and manual inspection, and the model is evaluated against baselines implemented from official sources or retrained on GES-Inter. Evaluating on a newly introduced dataset is a benchmark-construction choice, not an equation-level circularity. The paper's own limitation statement (Section 5: 'the automatic pose extraction stream may influence our dataset with some bad instances') and Appendix A.1.1's admission of unavoidable PyMAF-X inaccuracies are validity and generalization concerns about pseudo-label quality, but they do not make any prediction equivalent to an input by definition. The only self-citation (Qi et al., 2024c) is cited for the pseudo-label data-collection convention, which is not load-bearing for the architecture or the reported improvements. No circular step satisfying the quoted-evidence requirement was found.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central contribution rests on dataset quality assumptions rather than on new physical or mathematical entities. The model introduces no invented particles or forces; the key free choices are the loss weight, filtering thresholds, and the unvalidated accuracy of pseudo-labels and audio alignment.

free parameters (2)
  • lambda_simple = 15
    Trade-off weight for L_simple in Eq. (4), set empirically with no sensitivity analysis reported. Ablations test architectural components, not this weight.
  • pose filtering thresholds = wrist Euler angle >150 degrees rejected; inter-frame change >25 degrees rejected
    Chosen by hand to discard jittery poses during dataset construction; these thresholds shape the distribution of GES-Inter and are not justified by independent criteria.
assumptions (4)
  • domain assumption PyMAF-X pseudo-labels from monocular talk show videos are accurate enough to serve as ground truth for training and evaluation.
    Section 3.1 and Appendix A.1.1 rely on PyMAF-X for all pose annotations; no independent validation against motion capture is reported.
  • domain assumption Speaker diarization and manual audio-speaker alignment correctly assign each separated audio stream to the visible speaker (claimed 95-98% accuracy).
    Section 3.1 and Appendix A.1.2; if identity assignment is wrong, the audio-conditioned gesture pairs are misaligned.
  • standard math Standard diffusion denoising objective and cosine noise schedule produce valid gesture distributions.
    Eq. (3) and Section 4.1 follow Nichol & Dhariwal (2021); this is accepted background.
  • domain assumption FGD, BC, and Diversity metrics computed with a pre-trained autoencoder are meaningful for comparing gesture quality.
    Section 4.1; the autoencoder is not described or released, and BC is a heuristic rhythm measure.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Co$^{3}$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion." pith.science (2026). https://pith.science/paper/C5LQ7RQA

@misc{pith2026250501746,
  author       = {Pith},
  title        = {Pith review of: Co$^3$Gesture: Towards Coherent Concurrent Co-speech 3D Gesture Generation with Interactive Diffusion},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C5LQ7RQA}},
  note         = {Machine review of arXiv:2505.01746}
}
abstract

Generating gestures from human speech has gained tremendous progress in animating virtual avatars. While the existing methods enable synthesizing gestures cooperated by individual self-talking, they overlook the practicality of concurrent gesture modeling with two-person interactive conversations. Moreover, the lack of high-quality datasets with concurrent co-speech gestures also limits handling this issue. To fulfill this goal, we first construct a large-scale concurrent co-speech gesture dataset that contains more than 7M frames for diverse two-person interactive posture sequences, dubbed GES-Inter. Additionally, we propose Co$^3$Gesture, a novel framework that enables coherent concurrent co-speech gesture synthesis including two-person interactive movements. Considering the asymmetric body dynamics of two speakers, our framework is built upon two cooperative generation branches conditioned on separated speaker audio. Specifically, to enhance the coordination of human postures with respect to corresponding speaker audios while interacting with the conversational partner, we present a Temporal Interaction Module (TIM). TIM can effectively model the temporal association representation between two speakers' gesture sequences as interaction guidance and fuse it into the concurrent gesture generation. Then, we devise a mutual attention mechanism to further holistically boost learning dependencies of interacted concurrent motions, thereby enabling us to generate vivid and coherent gestures. Extensive experiments demonstrate that our method outperforms the state-of-the-art models on our newly collected GES-Inter dataset. The dataset and source code are publicly available at \href{https://mattie-e.github.io/Co3/}{\textit{https://mattie-e.github.io/Co3/}}.

Figures

Figures reproduced from arXiv: 2505.01746 by the authors.

Figure 1
Figure 1. Diverse exemplary clips sampled by our method from our newly collected [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of our audio separation and alignment with speakers. to obtain high-quality poses (i.e., SMPL￾X Pavlakos et al. (2019) and FLAME Li et al. (2017)) from in-the-wild talk show videos. To obtain the individual sound signals of each speaker in the conversation while preserving the identity consistency with the posture movement, we employ the pyannote-audio Bredin et al. (2020) to separate the mixed speech, … view at source ↗
Figure 3
Figure 3. The overall pipeline of our Co3Gesture . Given conversational speech audios, our framework generates concurrent co-speech gestures with coherent interactions. data processing to filter the unnatural and jittery poses, thereby ensuring the high-quality of the dataset*. Our GES-Inter includes more than 7M validated gesture frames with 70 hours. To the best of our knowledge, this is the first large-scale co-speech data… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Visualization of our generated concurrent 3D co-speech gestures against various state-of [PITH_FULL_IMAGE:figures/full_fig_p010_4.png]
Figure 5
Figure 5. Figure 5: User study on gesture naturalness, mo￾tion smoothness, and interaction coherency. Our framework demonstrates the best perfor￾mance compared with all the competitors. To be specific, our method achieves noticeable advan￾tages from the perspective of smoothness and inter…
Figure 6
Figure 6. Figure 6: The overall workflow of our dataset construction. The videos are processed to obtain [PITH_FULL_IMAGE:figures/full_fig_p016_6.png]
Figure 7
Figure 7. Figure 7: Visualization of our generated concurrent 3D co-speech gestures in the ablation study. [PITH_FULL_IMAGE:figures/full_fig_p019_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references · 13 canonical work pages

  1. [1]

    Style transfer for co-speech gesture animation: A multi-speaker conditional-mixture approach

    Chaitanya Ahuja, Dong Won Lee, Yukiko I Nakano, and Louis-Philippe Morency. Style transfer for co-speech gesture animation: A multi-speaker conditional-mixture approach. In Computer Vision–ECCV 2020: 16th European Conference, Glasgow, UK, August 23–28, 2020, Proceedings, Part XVIII 16, pp. 248–265. Springer,

    work page 2020

  2. [5]

    The genea challenge 2023: A large-scale evaluation of gesture generation models in monadic and dyadic settings

    Taras Kucherenko, Rajmund Nagy, Youngwoo Yoon, Jieyeon Woo, Teodor Nikolov, Mihail Tsakov, and Gustav Eje Henter. The genea challenge 2023: A large-scale evaluation of gesture generation models in monadic and dyadic settings. In Proceedings of the 25th International Conference on Multimodal Interaction, pp. 792–801,

    work page 2023

  3. [6]

    Gesture controllers

    Sergey Levine, Philipp Kr¨ahenb¨uhl, Sebastian Thrun, and Vladlen Koltun. Gesture controllers. In Acm siggraph 2010 papers, pp. 1–11

    work page 2010

  4. [9]

    Montreal forced aligner: Trainable text-speech alignment using kaldi

    Michael McAuliffe, Michaela Socolof, Sarah Mihuc, Michael Wagner, and Morgan Sonderegger. Montreal forced aligner: Trainable text-speech alignment using kaldi. In Interspeech, volume 2017, pp. 498–502,

    work page 2017

  5. [11]

    Emotiongesture: Audio- driven diverse emotional co-speech 3d gesture generation

    Xingqun Qi, Chen Liu, Lincheng Li, Jie Hou, Haoran Xin, and Xin Yu. Emotiongesture: Audio- driven diverse emotional co-speech 3d gesture generation. IEEE Transactions on Multimedia, 2024a. Xingqun Qi, Jiahao Pan, Peng Li, Ruibin Yuan, Xiaowei Chi, Mengfei Li, Wenhan Luo, Wei Xue, Shanghang Zhang, Qifeng Liu, and Yike Guo. Weakly-supervised emotion transit...

    arXiv

  6. [14]

    Pymaf-x: Towards well-aligned full-body model regression from monocular images

    Hongwen Zhang, Yating Tian, Yuxiang Zhang, Mengcheng Li, Liang An, Zhenan Sun, and Yebin Liu. Pymaf-x: Towards well-aligned full-body model regression from monocular images. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2023a. Jianrong Zhang, Yangsong Zhang, Xiaodong Cun, Yong Zhang, Hongwei Zhao, Hongtao Lu, Xi Shen, and Ying Shan. Gene...

    arXiv

  7. [15]

    15 Published as a conference paper at ICLR 2025 AutomaticProcessingManualProcessing Basic Video ProcessingAudio Extraction and Filtering Step

    work page 2025

  8. [16]

    They are then processed using automated methods to extract both audio and motion information

    A.1.1 A UTOMATIC PROCESSING STEPS To build a high-quality 3D co-speech gesture dataset with concurrent and interactive body dynamics, we collect a considerable number of videos. They are then processed using automated methods to extract both audio and motion information. Basic Video Processing (Step 1, 2, 3): First, with related searching keywords, we col...

    work page 2023

Show all 18 references
  1. [17]

    Due to the strict keyword selection in raw video crawling, our dataset rarely contains two speakers standing or walking around

    The jittery effects are mostly caused by the blurring of speakers moving quickly in consecutive video frames. Due to the strict keyword selection in raw video crawling, our dataset rarely contains two speakers standing or walking around. If there are several clips including th...

    2019

  2. [18]

    Meanwhile, our method displays a much lower standard deviation than InterX and InterGen

    Our method even achieves a 10% ((4.4-4.0)/4=10%) large marginal improvement over suboptimal InterGen in Naturalness. Meanwhile, our method displays a much lower standard deviation than InterX and InterGen. This indicates the much more stable performance of our method against c...

    2025

  3. [2010]

    Learning to generate diverse dance motions with transformer

    Jiaman Li, Yihang Yin, Hang Chu, Yi Zhou, Tingwu Wang, Sanja Fidler, and Hao Li. Learning to generate diverse dance motions with transformer. arXiv preprint arXiv:2008.08171,

    2008 arXiv

  4. [2017]

    Understanding embodied reference with touch-line transformer

    12 Published as a conference paper at ICLR 2025 Yang Li, Xiaoxue Chen, Hao Zhao, Jiangtao Gong, Guyue Zhou, Federico Rossano, and Yixin Zhu. Understanding embodied reference with touch-line transformer. In ICLR,

    2025

  5. [2019]

    Pyannote

    Herv´e Bredin, Ruiqing Yin, Juan Manuel Coria, Gregory Gelly, Pavel Korshunov, Marvin Lavechin, Diego Fustes, Hadrien Titeux, Wassim Bouaziz, and Marie-Philippe Gill. Pyannote. audio: neural building blocks for speaker diarization. In ICASSP 2020-2020 IEEE International Confer...

    2020

  6. [2020]

    Whisperx: Time-accurate speech transcription of long-form audio

    Max Bain, Jaesung Huh, Tengda Han, and Andrew Zisserman. Whisperx: Time-accurate speech transcription of long-form audio. INTERSPEECH 2023,

    2023

  7. [2021]

    Generative agents: Interactive simulacra of human behavior

    13 Published as a conference paper at ICLR 2025 Joon Sung Park, Joseph O’Brien, Carrie Jun Cai, Meredith Ringel Morris, Percy Liang, and Michael S Bernstein. Generative agents: Interactive simulacra of human behavior. In Proceedings of the 36th annual acm symposium on user int...

    2025

  8. [2022]

    Denoising diffusion implicit models

    Jiaming Song, Chenlin Meng, and Stefano Ermon. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502,

    2010 arXiv

  9. [2023]

    Alpacafarm: A simulation framework for methods that learn from human feedback

    11 Published as a conference paper at ICLR 2025 Yann Dubois, Chen Xuechen Li, Rohan Taori, Tianyi Zhang, Ishaan Gulrajani, Jimmy Ba, Carlos Guestrin, Percy S Liang, and Tatsunori B Hashimoto. Alpacafarm: A simulation framework for methods that learn from human feedback. Advanc...

    2025

  10. [2024]

    Diffusestylegesture: stylized audio-driven co-speech gesture generation with diffusion models

    14 Published as a conference paper at ICLR 2025 Sicheng Yang, Zhiyong Wu, Minglei Li, Zhensong Zhang, Lei Hao, Weihong Bao, Ming Cheng, and Long Xiao. Diffusestylegesture: stylized audio-driven co-speech gesture generation with diffusion models. In Proceedings of the Thirty-Se...

    2025

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.