Pith. sign in

REVIEW 4 major objections 5 minor 4 cited by

RealisDance-DiT: Simple yet Strong Baseline towards Controllable Character Animation in the Wild

T0 review · 4 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 The paper argues that a powerful video foundation model can be adapted to controllable character animation with minimal architectural edits, outperforming existing methods built on heavy reference networks.

desk verdict A plausible simple-recipe baseline for controllable character animation on a large DiT, but the headline causal claim about Reference Nets is not actually controlled. read the letter →

arxiv 2504.14977 v1 pith:OZO4CRT5 submitted 2025-04-21 cs.CV

Jingkai Zhou , Yifan Wu , Shikai Li , Min Wei , Chao Fan , Weihua Chen , Wei Jiang , Fan Wang This is my paper · ORCID
classification cs.CV
keywords controllablecharacteranimationvideodiffusionmodelDiTfoundationReferenceNetlow-noisewarmuplargebatchessmalliterationspose-guidedgenerationRealisDance-Val
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

RealisDance-DiT asks what happens when you stop adding heavy reference networks to a video diffusion model and start from a much larger pretrained video foundation model. The paper's answer is that a few simple changes — concatenating the pose and reference conditions into the DiT's latents, using a spatially shifted rotary position embedding for the reference latent, and fine-tuning only a subset of parameters — are enough for state-of-the-art controllable character animation. To keep the foundation model's priors intact, it proposes two fine-tuning tactics: a low-noise warmup that samples small diffusion timesteps early, and a "large batches, small iterations" schedule that limits the number of adaptation steps. Evaluated on a new open-world benchmark (RealisDance-Val) plus the TikTok and UBC fashion video datasets, the model reports the best or second-best score on most metrics, with the largest margins on FVD and FID. If the claim is right, the field should spend less effort on bypass networks and more on scaling and preserving strong video foundation models.

What carries the argument

The load-bearing mechanism is the minimal conditioning path into the DiT: pose latents from HaMeR, DWPose, and SMPL-CS are encoded by the Wan VAE and added to the noise latent, while the reference latent is concatenated along the sequence dimension, with its rotary position embedding spatially shifted to the first frame. This replaces the Reference Net with a small set of trainable components — condition patchifiers, a zero projection layer, and the self-attention blocks — so the foundation model's own priors do the work. The second mechanism is the fine-tuning recipe: the low-noise warmup defines a dynamic probability density $f(x)$ that samples small timesteps early and decays to uniform sampling after $\tau$ iterations, and "large batches, small iterations" keeps updates informative while limiting the number of adaptation steps to protect pretrained priors.

What would settle it

Evaluate the model on a set of extreme, heavily occluded poses (for example acrobatic or gymnastic sequences) where HaMeR, DWPose, and SMPL-CS are known to be inaccurate, and measure generated frames' pose alignment and artifact rate. If outputs remain clean and pose-accurate, the claim that simple modifications suffice in the wild is supported; if random poses and artifacts appear, the boundary described in the paper's limitation section is reached.

Watch

Extended reading notes

Core claim

The central discovery is that the widely adopted Reference Net — a separate branch that injects character identity into the generative model — is suboptimal for large DiT-based video foundation models like Wan-2.1. Minimal modifications, namely adding pose latents to the noise latent, concatenating the reference latent along the sequence dimension, and replacing the rotary position embedding with a spatially shifted version so the reference latent sits at the first frame, produce a stronger baseline than the Reference Net variants tested. The paper argues this works because the foundation model already contains the priors needed for open-world character consistency; fine-tuning's job is to unlock those priors, not to add structure. Two training strategies support this: a dynamic timestep-sampling distribution that favors low-noise samples early in fine-tuning, and a "large batches, small iterations" schedule that prevents downstream data from overwriting the pretrained prior. The resulting RealisDance-DiT is reported to outperform all compared methods by a large margin on FVD and FID on RealisDance-Val and to rank first or second across metrics on TikTok and UBC fashion video.

Load-bearing premise

The whole pipeline rests on the accuracy of three external pose estimators; when all three misestimate an extremely complex pose, the model cannot compensate and tends to generate random poses with artifacts, as the paper's own limitation section concedes.

Editorial extensions

If this is right

  • Reference Net branches become unnecessary for large DiT video foundation models; the concatenation-plus-shifted-RoPE design matches or beats them while adding far fewer parameters.
  • Fine-tuning only a subset of parameters (condition patchifiers, zero projection layer, self-attention blocks) is enough for downstream character animation, so the foundation model's priors are largely preserved.
  • Sampling low-noise examples early in fine-tuning accelerates convergence, while a high-noise warmup slows it down, indicating that early adaptation should avoid difficult samples.
  • Training with large batches and few iterations prevents overfitting to downstream data; too many iterations reduce output diversity and introduce background artifacts.
  • The new RealisDance-Val benchmark exposes open-world challenges — rare poses, stylized characters, object interactions, and complex lighting — that simpler benchmarks do not capture, and the proposed method performs best on the aggregate FVD and FID.

Reading between the lines

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

  • If RealisDance-DiT's success generalizes, future gains in character animation may come mainly from scaling and improving the video foundation model itself, rather than from inventing more elaborate conditioning architectures.
  • The method's dependence on HaMeR, DWPose, and SMPL-CS means its "in the wild" scope is bounded by pose-estimation quality; training with corrupted or noisy poses, or with a learned robust pose encoder, could push that boundary further than the paper's limitation section acknowledges.
  • The "large batches, small iterations" recipe is plausibly a general fine-tuning principle for other large generative models on narrow downstream tasks, where preserving pretrained priors matters more than minimizing the training loss.
  • The static-background failure in stationary-camera scenes suggests the model lacks explicit camera-motion or optical-flow conditioning; adding such cues could be a next step for cases where the pose sequence carries little motion information.
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

4 major / 5 minor

Summary. The paper proposes RealisDance-DiT, a controllable character animation method built on the Wan-2.1 I2V 14B video foundation model. The core claim is that with a sufficiently powerful foundation model, simple architectural modifications (concatenation of reference and pose latents, shifted RoPE) and two fine-tuning strategies (low-noise warmup timestep sampling and 'large batches, small iterations') can replace the heavy Reference Net designs used in prior methods. The authors introduce a new test set, RealisDance-Val, and report FVD/FID improvements over Animate-X, ControlNeXt, MimicMotion, MooreAA, and MusePose on TikTok, UBC Fashion, and RealisDance-Val. The paper also contains ablations on architecture and fine-tuning strategies, and a limitations section.

Significance. If the central claim is correct, this is a useful counterpoint to the prevailing Reference Net design for large DiT video models, and the proposed baseline plus the new benchmark would be a service to the community. The paper is clearly written and the qualitative results are visually compelling. Credit is due for explicitly acknowledging the dependency on three external pose estimators and for stating the 'large batches, small iterations' recipe. However, the empirical support for the headline claims is currently confounded by base-model scale and by an under-documented Reference Net ablation, so the significance can only be provisionally assessed.

major comments (4)
  1. [§4.1, Tables 1-3] The headline 'large margin' comparisons are confounded by base-model scale: RealisDance-DiT is built on Wan-2.1 I2V 14B, while Animate-X, ControlNeXt, MimicMotion, MooreAA, and MusePose are based on substantially smaller Stable Diffusion UNet backbones. The reported FVD/FID gains may therefore reflect the pre-training and capacity of the 14B base rather than the proposed architectural modifications or fine-tuning strategies. Please provide a matched-scale baseline, e.g., fine-tune Wan-2.1 I2V 14B with the same data but with a standard Reference Net or ControlNet-style injection, or at least report the performance of the unmodified Wan-2.1 I2V with pose conditioning via existing adapters, to isolate the contribution of the proposed recipe.
  2. [§4.2, Table 4] The conclusion that 'Reference Net is suboptimal for large DiT models' is not supported by the presented evidence. The full Reference Net variant could not be trained (OOM), and the 'Light Ref. Net' variant prunes the Reference Net to one block per five main-network blocks with no reported convergence criterion, number of training steps, or final loss. Its inferior FID (31.01) and FVD (678.98) may be due to under-training or capacity reduction rather than the Reference Net design itself. Please train a Reference Net variant at matched scale and compute budget (e.g., with gradient checkpointing, offloading, or a lighter full Reference Net) and report training curves and final loss. Also, the claim that part fine-tuning 'will not degrade the final performance' is internally contradicted: Part Ft improves FID (24.79 vs 25.58) but worsens FVD (563.28 vs 519.22); please discuss this trade-off or provide repeated runs with confidence intervals.
  3. [§3.2, Eqs. (2)-(3)] The low-noise warmup strategy depends on hyperparameters α and τ, but the paper does not report the values used for the final model, nor any sensitivity analysis. The ablation in Figure 8 is on the 1.3B T2V model and only shows smoothed loss curves, not the effect on FVD/FID or generation quality. Please quantify the impact of the proposed sampling schedule on the final metrics (e.g., uniform vs. low-noise warmup on the same base model) and report α, τ, and the total number of training iterations, along with the batch sizes used in the 'large batches and small iterations' strategy.
  4. [§5 (Limitations)] The paper acknowledges that when all three pose estimators fail, the model generates random poses and artifacts. This limitation is common to pose-conditioned methods, but it interacts with the central claim: the evaluation and the method's success depend on upstream pose quality. The paper should quantify how often pose estimates are inaccurate on the RealisDance-Val dataset and discuss the impact on the reported metrics, or at least include a robustness analysis to noisy pose inputs.
minor comments (5)
  1. [§3.1] The 'zero projection layer' is mentioned in the overview but never formally defined; please specify its role, placement, and initialization.
  2. [§4.1] There is a typo 'RealiDance-DiT' in the qualitative comparison paragraph; it should be 'RealisDance-DiT'.
  3. [§4.2, Figure 9] The text describes different batch configurations but does not state the actual batch sizes and iteration counts used to produce Figure 9; please provide these details.
  4. [§4.1, Table 3] On the UBC Fashion dataset, RealisDance-DiT ranks second on FVD and FID behind Animate-X; the phrase 'large margin' in the abstract and conclusion should be qualified to reflect that the large gains appear only on the proposed RealisDance-Val benchmark.
  5. [§4 (Dataset)] The RealisDance-Val dataset description (100 videos) lacks basic statistics such as resolution distribution, number of characters, scene types, and pose complexity measures; please include a summary table or histogram.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the paper's claims are empirical, and its ablations and external benchmarks do not reduce to fitted inputs or self-citations.

full rationale

This is an empirical systems paper rather than a derivational one. The central claims—that simple architecture modifications suffice for a strong DiT foundation model, that low-noise warmup accelerates convergence, and that large batches with fewer iterations preserve priors—are supported by controlled ablations (Table 4, Figures 8-9) and by external benchmarks (TikTok and UBC fashion video datasets, Tables 2-3). The low-noise warmup schedule in Eqs. (2)-(3) introduces hyperparameters alpha and tau, but these are not fitted to test targets and the schedule is not a prediction derived from the evaluation metrics. The only self-citation is the reuse of the pose-condition pipeline from RealisDance [35] (HaMeR, DWPose, SMPL-CS), which is an input component, not a load-bearing derivation of the paper's conclusions. The RealisDance-Val dataset is self-collected, but the paper also evaluates on established external datasets, so the central comparison is not equivalent to the paper's own inputs by construction. No equation, architecture choice, or fitted parameter is shown to be definitionally tied to the reported FVD/FID values, so no circular step is present.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the Wan-2.1 foundation model's priors, the accuracy of external pose estimators, the validity of the self-curated benchmark and metrics, and the quality and non-overlap of vendor-purchased training data. Hand-chosen hyperparameters alpha and tau control the low-noise warmup schedule; the learning rate is fixed at 1e-5. No new entities are introduced.

free parameters (3)
  • low-noise warmup strength alpha = in [0,1], value not stated
    Hyperparameter controlling the maximum deviation from uniform timestep sampling in Eq. (3); chosen by hand, not fitted to a validation set.
  • warmup threshold tau = not stated
    Number of iterations during which low-noise sampling is applied (Eq. 3); hand-chosen.
  • learning rate = 1e-5
    Optimizer hyperparameter for AdamW used in all fine-tuning runs; chosen by hand.
assumptions (4)
  • domain assumption Wan-2.1 foundation model weights are a valid starting point and its pre-trained priors are beneficial for downstream character animation.
    Section 3.1 builds RealisDance-DiT on Wan-2.1; the claim that the model 'inherently possesses the capability' is assumed rather than proven.
  • domain assumption The three pose conditions (HaMeR, DWPose, SMPL-CS) provide sufficiently accurate driving signals for animation.
    Section 3.1 uses these pose latents as input; Section 5 concedes failure when all three estimates are wrong.
  • domain assumption VBench-I2V metrics and the self-curated RealisDance-Val set are valid measures of open-world animation quality.
    Section 4.1 justifies rejecting low-level metrics post hoc and relies on VBench-I2V plus FVD/FID for the main comparison.
  • domain assumption The 1M vendor-purchased videos are high-quality and non-overlapping with the test sets.
    Section 4 states that both training sets exclude the test datasets, but the provenance is vendor data and cannot be independently verified.

how reviews work

0 comments
Cite this review

Pith. "Pith review of RealisDance-DiT: Simple yet Strong Baseline towards Controllable Character Animation in the Wild." pith.science (2026). https://pith.science/paper/OZO4CRT5

@misc{pith2026250414977,
  author       = {Pith},
  title        = {Pith review of: RealisDance-DiT: Simple yet Strong Baseline towards Controllable Character Animation in the Wild},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OZO4CRT5}},
  note         = {Machine review of arXiv:2504.14977}
}
read the original abstract

Controllable character animation remains a challenging problem, particularly in handling rare poses, stylized characters, character-object interactions, complex illumination, and dynamic scenes. To tackle these issues, prior work has largely focused on injecting pose and appearance guidance via elaborate bypass networks, but often struggles to generalize to open-world scenarios. In this paper, we propose a new perspective that, as long as the foundation model is powerful enough, straightforward model modifications with flexible fine-tuning strategies can largely address the above challenges, taking a step towards controllable character animation in the wild. Specifically, we introduce RealisDance-DiT, built upon the Wan-2.1 video foundation model. Our sufficient analysis reveals that the widely adopted Reference Net design is suboptimal for large-scale DiT models. Instead, we demonstrate that minimal modifications to the foundation model architecture yield a surprisingly strong baseline. We further propose the low-noise warmup and "large batches and small iterations" strategies to accelerate model convergence during fine-tuning while maximally preserving the priors of the foundation model. In addition, we introduce a new test dataset that captures diverse real-world challenges, complementing existing benchmarks such as TikTok dataset and UBC fashion video dataset, to comprehensively evaluate the proposed method. Extensive experiments show that RealisDance-DiT outperforms existing methods by a large margin.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 4 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. UniMoCa: Unifying Motion and Camera Controls as Visual Proxies for Faithful Human Video Generation

    cs.CV 2026-08 conditional novelty 6.0 of 10

    A visual proxy that renders human motion under the driving camera and overlays camera trajectory markers lets a video diffusion model control both body motion and camera movement from a single visual conditioning space.

  2. 3D Scene-Adaptive Trajectory-Controllable Human Image Animation with Camera Movement

    cs.CV 2026-06 unverdicted novelty 6.0 of 10

    Presents a scene-adaptive 3D human image animation framework using ground-adaptive motion retargeting and viewpoint-adaptive latent fusion to control human and camera trajectories, claiming improvements on two benchmarks.

  3. AHOY! Animatable Humans under Occlusion from YouTube Videos with Gaussian Splatting and Video Diffusion Priors

    cs.CV 2026-03 conditional novelty 6.0 of 10

    Identity-finetuned video diffusion plus RF-Inversion can supply multi-view body supervision that lets 3D Gaussian avatars be completed and animated from heavily occluded monocular video.

  4. Extension of generalized KYP lemma: from LTI systems to LPV systems

    math.DS 2025-08 unverdicted novelty 6.0 of 10

    The abstract claims a gKYP lemma extension for LPV systems via frequency-range enlargement, but the submitted full text is an unrelated video generation paper.

Reference graph

Works this paper leans on

36 extracted references · 21 canonical work pages · cited by 4 Pith papers

  1. [1]

    https://github.com/MooreThreads/Moore- 9 AnimateAnyone, 2024

    Moore-animateanyone. https://github.com/MooreThreads/Moore- 9 AnimateAnyone, 2024

    work page 2024

  2. [2]

    https://viggleai.io/, 2024

    Viggle ai. https://viggleai.io/, 2024

    2024

  3. [3]

    Large scale GAN training for high fidelity natural image synthesis

    Andrew Brock, Jeff Donahue, and Karen Si- monyan. Large scale GAN training for high fidelity natural image synthesis. In7th Interna- tional Conference on Learning Representations, 2019

    work page 2019

  4. [4]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, et al. Scaling rectified flow transformers for high-resolution image synthesis. In Forty-first International Conference on Machine Learning, 2024

    work page 2024

  5. [5]

    Humandit: Pose- guided diffusion transformer for long-form hu- man motion video generation

    Qijun Gan, Yi Ren, Chen Zhang, Zhenhui Ye, Pan Xie, Xiang Yin, Zehuan Yuan, Bingyue Peng, and Jianke Zhu. Humandit: Pose- guided diffusion transformer for long-form hu- man motion video generation. arXiv preprint arXiv:2502.04847, 2025

    arXiv 2025

  6. [6]

    Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C

    Ian J. Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, David Warde-Farley, Sherjil Ozair, Aaron C. Courville, and Yoshua Bengio. Generative adversarial nets. InAdvances in Neu- ral Information Processing Systems, volume 27, pages 2672–2680, 2014

    work page 2014

  7. [7]

    Efficient diffusion training via min-snr weighting strategy

    Tiankai Hang, Shuyang Gu, Chen Li, Jianmin Bao, Dong Chen, Han Hu, Xin Geng, and Bain- ing Guo. Efficient diffusion training via min-snr weighting strategy. InIEEE/CVF International Conference on Computer Vision, ICCV 2023, pages 7407–7417, 2023

    work page 2023

  8. [8]

    Classifier- free diffusion guidance

    Jonathan Ho and Tim Salimans. Classifier- free diffusion guidance. arXiv preprint arXiv:2207.12598, 2022

    arXiv 2022

Show all 36 references
  1. [9]

    Animate anyone: Con- sistent and controllable image-to-video synthesis for character animation

    Li Hu, Xin Gao, Peng Zhang, Ke Sun, Bang Zhang, and Liefeng Bo. Animate anyone: Con- sistent and controllable image-to-video synthesis for character animation. InIEEE/CVF Confer- ence on Computer Vision and Pattern Recogni- tion, pages 8153–8163, 2024

    2024

  2. [10]

    Animate any- one 2: High-fidelity character image animation with environment affordance

    Li Hu, Guangyuan Wang, Zhen Shen, Xin Gao, Dechao Meng, Lian Zhuo, Peng Zhang, Bang Zhang, and Liefeng Bo. Animate any- one 2: High-fidelity character image animation with environment affordance. arXiv preprint arXiv:2502.06145, 2025

    2025 arXiv

  3. [11]

    VBench++: Comprehensive and versatile benchmark suite for video generative models

    Ziqi Huang, Fan Zhang, Xiaojie Xu, Yinan He, Jiashuo Yu, Ziyue Dong, Qianli Ma, Nattapol Chanpaisit, Chenyang Si, Yuming Jiang, Yao- hui Wang, Xinyuan Chen, Ying-Cong Chen, Limin Wang, Dahua Lin, Yu Qiao, and Ziwei Liu. VBench++: Comprehensive and versatile benchmark suite for...

    2024 arXiv

  4. [12]

    Learning high fidelity depths of dressed humans by watch- ing social media dance videos

    Yasamin Jafarian and Hyun Soo Park. Learning high fidelity depths of dressed humans by watch- ing social media dance videos. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 12753–12762, 2021

    2021

  5. [13]

    Vace: All-in- one video creation and editing.arXiv preprint arXiv:2503.07598, 2025

    Zeyinzi Jiang, Zhen Han, Chaojie Mao, Jingfeng Zhang, Yulin Pan, and Yu Liu. Vace: All-in- one video creation and editing.arXiv preprint arXiv:2503.07598, 2025

    2025 arXiv

  6. [14]

    Immiscible diffusion: Accelerating diffu- sion training with noise assignment

    Yiheng Li, Heyang Jiang, Akio Kodaira, Masayoshi Tomizuka, Kurt Keutzer, and Chen- feng Xu. Immiscible diffusion: Accelerating diffu- sion training with noise assignment. InAdvances in Neural Information Processing Systems 38: Annual Conference on Neural Information Pro- cessi...

    2024

  7. [15]

    Omnihuman-1: Rethinking the scaling-up of one-stage condi- tioned human animation models.arXiv preprint arXiv:2502.01061, 2025

    Gaojie Lin, Jianwen Jiang, Jiaqi Yang, Ze- rong Zheng, and Chao Liang. Omnihuman-1: Rethinking the scaling-up of one-stage condi- tioned human animation models.arXiv preprint arXiv:2502.01061, 2025

    2025 arXiv

  8. [16]

    Georgios Pavlakos, Vasileios Choutas, Nima Ghorbani, Timo Bolkart, Ahmed A. A. Osman, Dimitrios Tzionas, and Michael J. Black. Ex- pressive body capture: 3d hands, face, and body from a single image. InProceedings IEEE Conf. on Computer Vision and Pattern Recognition (CVPR), 2019

    2019

  9. [17]

    Reconstructing hands in 3d with transformers

    Georgios Pavlakos, Dandan Shan, Ilija Ra- dosavovic, Angjoo Kanazawa, David Fouhey, and Jitendra Malik. Reconstructing hands in 3d with transformers. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024, pages 9826–9836, 2024

    2024

  10. [18]

    Con- trolnext: Powerful and efficient control for image and video generation

    Bohao Peng, Jian Wang, Yuechen Zhang, Wenbo Li, Ming-Chang Yang, and Jiaya Jia. Con- trolnext: Powerful and efficient control for image and video generation. arXiv preprint arXiv:2408.06070, 2024

    2024 arXiv

  11. [19]

    Learning transferable visual models from natural language supervision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agar- wal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. Learning transferable visual models from natural language supervision. In Marina Meila an...

    2021

  12. [20]

    World-grounded human motion recovery via gravity-view coordinates

    Zehong Shen, Huaijin Pi, Yan Xia, Zhi Cen, Sida Peng, Zechen Hu, Hujun Bao, Ruizhen Hu, and Xiaowei Zhou. World-grounded human motion recovery via gravity-view coordinates. In SIGGRAPH Asia Conference Proceedings, 2024

    2024

  13. [21]

    First order motion model for image animation

    Aliaksandr Siarohin, Stéphane Lathuilière, Sergey Tulyakov, Elisa Ricci, and Nicu Sebe. First order motion model for image animation. In Advances in Neural Information Processing Systems, volume 32, pages 7135–7145, 2019

    2019

  14. [22]

    Woodford, Jian Ren, Menglei Chai, and Sergey Tulyakov

    Aliaksandr Siarohin, Oliver J. Woodford, Jian Ren, Menglei Chai, and Sergey Tulyakov. Mo- tionrepresentationsforarticulatedanimation. In IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 13653–13662, 2021

    2021

  15. [23]

    Jianlin Su, Murtadha H. M. Ahmed, Yu Lu, Shengfeng Pan, Wen Bo, and Yunfeng Liu. Ro- former: Enhanced transformer with rotary po- sition embedding. Neurocomputing, 568:127063, 2024

    2024

  16. [24]

    Animate-x: Universal character image anima- tion with enhanced motion representation.arXiv preprint arXiv:2410.10306, 2024

    Shuai Tan, Biao Gong, Xiang Wang, Shi- wei Zhang, Dandan Zheng, Ruobing Zheng, Kecheng Zheng, Jingdong Chen, and Ming Yang. Animate-x: Universal character image anima- tion with enhanced motion representation.arXiv preprint arXiv:2410.10306, 2024

    arXiv

  17. [25]

    Musepose: a pose- driven image-to-video framework for virtual hu- man generation

    Zhengyan Tong, Chao Li, Zhaokang Chen, Bin Wu, and Wenjiang Zhou. Musepose: a pose- driven image-to-video framework for virtual hu- man generation. arxiv, 2024

    2024

  18. [26]

    Acloserlookattime steps is worthy of triple speed-up for diffusion model training

    Kai Wang, Yukun Zhou, Mingjia Shi, Zhihang Yuan, Yuzhang Shang, Xiaojiang Peng, Han- wangZhang, andYangYou. Acloserlookattime steps is worthy of triple speed-up for diffusion model training. arXiv preprint arXiv:2405.17403, 2024

    2024 arXiv

  19. [27]

    Disco: Disentangled control for realistic human dance generation

    Tan Wang, Linjie Li, Kevin Lin, Yuanhao Zhai, Chung-Ching Lin, Zhengyuan Yang, Hanwang Zhang, Zicheng Liu, and Lijuan Wang. Disco: Disentangled control for realistic human dance generation. In IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 9326–9336, 2024

    2024

  20. [28]

    Humanvid: Demystifying training data for camera-controllable human image animation

    Zhenzhi Wang, Yixuan Li, Yanhong Zeng, Youqing Fang, Yuwei Guo, Wenran Liu, Jing Tan, Kai Chen, Tianfan Xue, Bo Dai, and Dahua Lin. Humanvid: Demystifying training data for camera-controllable human image animation. In The Thirty-eight Conference on Neural Infor- mation Proces...

    2024

  21. [29]

    Wan: Open and advanced large-scale video generative models

    WanTeam, :, Ang Wang, Baole Ai, Bin Wen, Chaojie Mao, Chen-Wei Xie, Di Chen, Feiwu Yu, HaimingZhao, JianxiaoYang, JianyuanZeng, Ji- ayu Wang, Jingfeng Zhang, Jingren Zhou, Jinkai Wang, Jixuan Chen, Kai Zhu, Kang Zhao, Keyu Yan, Lianghua Huang, Mengyang Feng, Ningyi Zhang, Pand...

    2025 arXiv

  22. [30]

    Magicanimate: Temporally consistent human image animation using diffusion model

    Zhongcong Xu, Jianfeng Zhang, Jun Hao Liew, Hanshu Yan, Jia-Wei Liu, Chenxu Zhang, Jiashi Feng, and Mike Zheng Shou. Magicanimate: Temporally consistent human image animation using diffusion model. InIEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1481–1490, 2024

    2024

  23. [31]

    Effectivewhole-bodyposeestimationwith two-stages distillation

    Zhendong Yang, Ailing Zeng, Chun Yuan, and YuLi. Effectivewhole-bodyposeestimationwith two-stages distillation. InIEEE/CVF Interna- tional Conference on Computer Vision, pages 4212–4222, 2023

    2023

  24. [32]

    Dwnet: Dense warp-based network for pose-guided human video generation

    Polina Zablotskaia, Aliaksandr Siarohin, Bo Zhao, and Leonid Sigal. Dwnet: Dense warp-based network for pose-guided human video generation. In 30th British Machine Vision Conference, page 51, 2019

    2019

  25. [33]

    Adding conditional control to text-to-image dif- fusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image dif- fusion models. InIEEE/CVF International Con- ference on Computer Vision, pages 3813–3824, 2023

    2023

  26. [34]

    Mimicmotion: High-quality human mo- tion video generation with confidence-aware pose guidance

    Yuang Zhang, Jiaxi Gu, Li-Wen Wang, Han Wang, Junqi Cheng, Yuefeng Zhu, and Fangyuan Zou. Mimicmotion: High-quality human mo- tion video generation with confidence-aware pose guidance. arXiv preprint arXiv:2406.19680, 2024. 11

    2024 arXiv

  27. [35]

    Realisdance: Equip controllable character animation with re- alistic hands

    Jingkai Zhou, Benzhi Wang, Weihua Chen, Jingqi Bai, Dongyang Li, Aixi Zhang, Hao Xu, Mingyang Yang, and Fan Wang. Realisdance: Equip controllable character animation with re- alistic hands. arXiv preprint arXiv:2409.06202, 2024

    2024 arXiv

  28. [36]

    Champ: Controllable and consistent human image animation with 3d parametric guidance

    Shenhao Zhu, Junming Leo Chen, Zuozhuo Dai, Zilong Dong, Yinghui Xu, Xun Cao, Yao Yao, Hao Zhu, and Siyu Zhu. Champ: Controllable and consistent human image animation with 3d parametric guidance. InEuropean Conference on Computer Vision, volume 15113, pages 145–162, 2024. 12

    2024

Pith tools

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