Sketch2Anim: Towards Transferring Sketch Storyboards into 3D Animation

Changjian Li; Chuan Guo; Jiawei Wang; Lei Zhong; Yiming Xie

REVIEW 4 major objections 6 minor 10 references

Sketch2Anim: Towards Transferring Sketch Storyboards into 3D Animation

T0 review · 4 major / 6 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 This paper claims a first: a sketch storyboard — stick-figure keypose, traced joint trajectories, and an action word — can drive a motion-diffusion model to generate a 3D animation clip directly, with no manual keyframing and no 3D…

desk verdict Solid systems paper: the 2D-3D embedding mapper and keypose adapter are real contributions, but the end-to-end storyboard claim rests mainly on qualitative evidence because the quantitative evaluation uses synthetic projections. read the letter →

arxiv 2504.19189 v1 pith:B2GJLH5Z submitted 2025-04-27 cs.GR cs.CV

Lei Zhong , Chuan Guo , Yiming Xie , Jiawei Wang , Changjian Li This is my paper · ORCID

classification cs.GRcs.CV

keywords sketch-basedanimationstoryboardmotiondiffusionmodelconditionalsynthesiskeyposecontroltrajectory2D-3Dembeddingalignmentediting

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

Sketch2Anim claims to be the first system that transfers a 2D sketch storyboard into 3D animation by treating the storyboard as three conditioning signals: a stick-figure keypose, joint-trajectory strokes, and an action word. The payoff, if the claim holds, is that animators can prototype motion by drawing rather than by manually posing joints in 3D software through trial and error. The paper's core trick is to avoid lifting 2D drawings into 3D altogether: it trains a motion generator on precise 3D keyposes and trajectories, then trains a neural mapper that aligns 2D sketch embeddings with those 3D conditions in a shared embedding space, so inference can consume 2D drawings directly. Experiments on synthetic projections of a motion dataset and on real user sketches, plus a perceptual study, back the claim.

What carries the argument

The load-bearing machinery is the 2D-3D neural mapper plus the two-module conditioning stack. The neural mapper is a pair of 2D encoders trained to push 2D keyposes and trajectories into the same embedding space as the frozen 3D encoders, using a matching loss, a CLIP-style contrastive loss, and a noise-reconstruction loss that couples alignment to generation. The conditioning stack pairs a trajectory ControlNet — a trainable copy of the diffusion model's transformer encoder whose residual features steer the frozen denoiser — with a trajectory-aware keypose adapter, a second trainable copy that consumes the ControlNet's residuals plus a grounded action embedding $\mathbf{a}' = \mathbf{a} + E^{3D}_k(K^{3D})$ and outputs corrective residuals. This division lets the trajectory control carry global dynamics while the adapter pins the local pose at the keypose timestep, which the ablation shows beats both a single fused ControlNet and two parallel ControlNets.

What would settle it

Run the full pipeline on a corpus of authentic hand-drawn storyboards spanning varied styles and drawing quality, and measure the failure rate of Sketch2Pose joint detection and the resulting motion quality against the Table 1 numbers; the end-to-end claim fails if detection errors on genuine cartoons are frequent or if quality drops substantially relative to the synthetic-projection evaluation.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes a conditional motion-generation recipe: a pre-trained latent motion diffusion model, frozen, is steered by two lightweight trainable modules — a trajectory ControlNet that injects joint-trajectory residuals, and a trajectory-aware keypose adapter that refines those residuals with the keypose condition. Because the keypose is treated as a grounding of the action word (its embedding is added to the text embedding), the adapter adds local pose constraints without disturbing the trajectory's global dynamics. A second training stage aligns 2D and 3D keypose and trajectory encoders in a shared embedding space using a matching loss, a noise-reconstruction loss, and a contrastive loss, so that at inference the user's 2D sketch lands exactly where the 3D conditions used in training do. The result is the claimed first direct 2D-storyboard-to-3D-motion transfer, with reported trajectory errors roughly halved and FID roughly five times better than training the generator directly on 2D conditions.

Load-bearing premise

The load-bearing premise is that Sketch2Pose will reliably find the 2D joints in whatever the user draws, and that orthographically projected motion-capture data with camera, joint, and body-proportion perturbations adequately stands in for real storyboard drawings — the authors themselves concede that joint detection "might fail if the drawing deviates too much from a reasonable human character."

Editorial extensions

If this is right

An animator can go from sketch to motion clip in about half a second of generation time: a 40-frame clip conditioned on detected 2D joints, traced trajectories, and the action word, with no manual keyframing.
Generated clips become directly editable in 3D: dragging keypose or trajectory points produces an updated motion, because the generator consumes these conditions natively.
Separating trajectory dynamics from keypose statics — rather than fusing both into one control signal — is what the ablations attribute the control gains to.
Storyboard frames compose into a single coherent animation through inversion-based blending, making the pipeline operate on whole storyboards, not just isolated frames.
On the reported benchmarks, 2D-3D embedding alignment beats direct 2D-conditioned training by roughly fivefold in FID and halves trajectory error, and beats explicit 2D-to-3D lifting on every reported metric.

Reading between the lines

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

The train-on-3D-surrogates, infer-on-2D pattern is a general recipe: any diffusion generator with well-defined 3D conditions could gain a sketch interface the same way, from scene layout to character-object interaction, without retraining the base model.
The reported asymmetry — keyposes converge nearly ten times faster than trajectories — suggests a design principle that likely extends beyond this paper: when fusing controls with very different learning speeds, layer the slower, static condition as a refinement of the faster, dynamic one's residuals.
Because the quantitative evaluation uses synthetic orthographic projections rather than authentic cartoons, feeding real hand-drawn storyboards through the mapper during alignment training, using detection confidence as a per-joint weight, is a plausible route to closing the domain gap the authors acknowledge.
Retrieval nearly tied generation on perceived realism in the user study, so a hybrid that retrieves a close motion and then regenerates it under the sketch constraints might combine retrieval's naturalness with the generator's controllability.

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.

Desk Editor's Note

read the letter

What to know: this paper does something new and mostly works. It is the first to take the full combination of 2D sketch keyposes, joint trajectory strokes, and action words and produce 3D motion. The architecture is well motivated: instead of fusing two ControlNets, they add a keypose adapter that refines the trajectory ControlNet's residual features, and they align 2D and 3D conditions in a shared embedding space so training can use precise 3D surrogates while inference consumes 2D sketches. The ablations genuinely support these choices over single and double ControlNet alternatives, and the code, data, and models are promised. The user study is a step beyond what this subfield usually ships.

The soft spots are mostly in the evaluation, and the paper is fairly honest about them. The quantitative benchmark is built by orthographically projecting HumanML3D motions, selecting a TMR-scored keypose, and adding camera, joint-noise, and body-proportion perturbations. That makes every test 2D signal exactly realizable by an in-distribution 3D motion, which is the favorable case for the approach. Real storyboard sketches appear only in qualitative figures and in the user study; there is no reported success rate or error distribution for Sketch2Pose on real drawings. The authors explicitly concede in Sec. 7.4 that Sketch2Pose may fail on sketches that deviate from a reasonable human figure. So the end-to-end claim "successfully transfers storyboards" is not quantitatively pinned down for actual target inputs.

A second concern: the trajectory control metric is also the objective of the L-BFGS inference guidance applied at test time (Eq. 15, Sec. C), so the headline Avg. Err. numbers partly reflect optimizer convergence. That said, Table A3 shows the method without inference guidance still beats all baselines on trajectory error, so the comparison is not hollow. Still, all tables are single-run point estimates with no error bars or significance tests, which is common in this literature but worth flagging. The user study reports 58 participants; motion-realism preference over retrieval is only 69%, which reads as an honest result.

Overall, the technical core is sound and the limitations are stated rather than hidden. The paper is a good TOG-style systems contribution. I would send it to review, and I would ask the authors for a real-sketch quantitative evaluation (even a modest one) and for error bars or significance tests on the main comparisons, plus a clearer separation of learned control from inference-guidance optimization. For someone working on sketch-based animation or conditional motion synthesis, this is worth a close read.

Referee Report

4 major / 6 minor

Summary. The paper proposes Sketch2Anim, a system for generating 3D human motion clips from 2D storyboard frames that contain a stick-figure keypose, optional joint-trajectory strokes, and an action word. The architecture has two modules: a multi-conditional latent diffusion motion generator, which combines a pre-trained motion diffusion model with a trajectory ControlNet and a trajectory-aware keypose adapter; and a neural mapper that aligns 2D and 3D keypose/trajectory embeddings so that 2D sketch conditions can be fed into the generator at inference time. The authors evaluate on the HumanML3D dataset with synthetic projection-based benchmarks, compare against three baselines, run ablations of the generator design and loss terms, and report a pairwise user study. They also demonstrate qualitative results on real hand-drawn storyboards and show a 3D editing application. The paper claims to be the first approach to adapt a motion diffusion model to generate 3D animations directly from 2D storyboards.

Significance. If the claims hold, the system is a practical contribution to sketch-based animation: it reduces manual keyframing by letting users specify a keypose, trajectory strokes, and an action word. The trajectory-aware keypose adapter is a reasonable design for combining multiple conditions in a pre-trained motion diffusion model, and the paper provides a substantial set of ablations (network module variants, loss terms, inference-guidance order), a user study, and releases code, data, and trained models. The central technical claims are, however, undermined by an apparent sign error in the alignment loss (Eq. 10), by the fact that the headline trajectory-control metric is the direct objective of a test-time L-BFGS optimization, and by the absence of a quantitative evaluation on real user sketches. These issues do not necessarily invalidate the system, but they must be addressed before the paper's broader claims can be accepted.

major comments (4)

[Sec. 5, Eq. (10)] The alignment loss is written as Lmatch = -1/B * sum ||s3D - s2D||^2 and then included as a positive term in Lalign (Eq. 13), which is minimized during training. This drives the paired embeddings apart rather than enforcing them to be close, contradicting the sentence immediately before Eq. (10) that says the goal is to make paired embeddings 'as close as possible.' Please correct the sign (or clarify if the loss is instead maximized) and state explicitly which form was actually used in training, since this is the core mechanism of the 2D-3D neural mapper.
[Sec. 6 and Suppl. C, Eqs. (15)-(16) and Table A3] The sharp improvement in Avg. Err.-2D and Avg. Err.-3D reported in Table 1 comes largely from a test-time L-BFGS inference guidance whose objective G(z_t, t, v, Tr_2D) is exactly the projected-trajectory error used as the evaluation metric. Table A3 shows that without inference guidance, Avg. Err.-2D is 0.158 (Average) and 0.159 (Cross), whereas with second-order guidance it drops to 0.0867 and 0.0792. The main comparison should report both guided and unguided numbers in the main text, and the text should be clear that the trajectory-accuracy numbers reflect a post hoc optimizer, not the learned controller alone. As written, the claim that the model 'effectively adheres to both keypose and joint trajectory constraints' is overstated for the unguided model.
[Sec. 7, Dataset processing, and Sec. 7.4 Limitations] The quantitative benchmark is constructed by orthographically projecting HumanML3D motions, selecting a TMR-scored keypose, and applying camera, joint-noise, and body-proportion augmentations. This means every test 2D signal is, by construction, a valid projection of an in-distribution 3D motion. Real user sketches appear only in qualitative results (Figs. 1, 6, 11, 12), and while the paper shows robust Sketch2Pose detection in those examples, Sec. 7.4 explicitly concedes that Sketch2Pose 'might fail if the drawing deviates too much from a reasonable human character.' No real-sketch success rate, joint-detection error distribution, or end-to-end quantitative result is reported. To support the abstract's claim that the system 'successfully transfers storyboards into high-quality 3D motions,' the authors should provide a quantitative evaluation on authentic user drawings, or alternatively narrow the claim to inputs that pass the current Sketch2Pose front end.
[Tables 1-3 and Fig. 10] All quantitative results are single-run point estimates with no error bars, confidence intervals, or significance tests. Given that diffusion sampling is stochastic and that several comparative differences are modest (e.g., Cross-setting FID 0.577 for Ours vs. 0.738 for Lift-and-Control, R-precision 0.815 vs. 0.796 in Table A2), the strength of the cross-method and cross-ablation conclusions is not statistically established. The authors should report multiple random seeds (or similar) together with variance, at least for the headline comparisons.

minor comments (6)

[Sec. 4.3, Eq. (8)] Eq. (8) uses the mask notation m'_i in both numerator and denominator, but the text defines only m_i. Please align the notation.
[Sec. 4.2, title] The subsection title contains a typo: 'Trajetory-aware Keypose Adapter' should be 'Trajectory-aware Keypose Adapter.'
[Sec. 4.2, Eq. (4)] The symbol Z(r') appears in Eq. (4) but Z is not defined in the main text at that point; a brief definition (zero-initialized linear layer) is given for the ControlNet in Sec. 4.1 but should be restated or referenced here for the adapter.
[Table A3 and Table 1] The 'Ours' row in Table A3 differs from the 'Ours' row in Table 1 because inference guidance is disabled in A3. This is stated in the supplementary text, but it would help readers if the main-text caption also noted that Table 1 includes inference guidance.
[Sec. 7.1, Motion Retrieval baseline] For the Motion Retrieval baseline, the FID is computed on retrieved real motions rather than on generated samples, which makes it not directly comparable to the generative FID numbers of the other methods. Please clarify this in the comparison description.
[Abstract and Sec. 1] The phrase 'we invent a neural mapper' is unusual for a research paper; consider 'we propose' or 'we design' instead.

Circularity Check

1 steps flagged · score 4.0 of 10

Reported Avg. Err.-2D in Table 1 is the exact objective minimized by L-BFGS inference guidance (Supp. Eqs. 15-16), so the trajectory-control number is partly optimizer convergence; no other derivation step reduces to its inputs, and same-author citations are not load-bearing.

self definitional [Sec. 6 (Inference Motion Generation); Supp. C Eqs. (15)-(16); Table 1 'Avg. Err.-2D']
"we employ inference guidance to enhance the accuracy of following 2D joint trajectories. The idea is to minimize the discrepancy between the projected trajectories of the generated motion and the 2D trajectories. ... epsilon_theta(z_t,t,a,T^r_2D,K_2D) = epsilon_theta(z_t,t,a,T^r_2D,K_2D) - tau_2 * H^{-1} grad_{z_t} G(z_t,t,v,T^r_2D), (15) G(z_t,t,v,T^r_2D) = sum_{i,j} m_ij ||P(R(hat x_0)_ij,v) - T^r_2D||^2_2 / sum_{i,j} m_ij (16)"

The headline trajectory-control metric Avg. Err.-2D is computed as the mean L2 distance between the projected generated motion and the input 2D trajectory (Sec. 7, 'Evaluation metrics'). Equation 16 defines the inference-guidance objective G as exactly this same quantity, and Eq. 15 takes L-BFGS steps on grad_{z_t} G at each denoising step (K=4 iterations). Therefore the Table 1 'Avg. Err.-2D' value partly reports convergence of the optimizer on its own objective, rather than how well the learned 2D-to-3D generation generalizes. The paper itself labels guidance a 'post-optimization process' in Supp. A. This makes the reported trajectory-adherence number partially true by construction, not by derivation. The keypose, text-motion, and realism metrics are not affected.

full rationale

Aside from the trajectory metric, the derivation chain is largely self-contained. The 3D conditional motion generator is trained with a noise-reconstruction loss plus explicit keypose and trajectory losses on HumanML3D (Eqs. 5-9), and the 2D-3D mapper is trained with matching, reconstruction, and contrastive losses (Eqs. 10-13); none of these define the output as the input. The 2D conditions used in training are orthographic projections of 3D motions, not the generated motions themselves. Same-author citations (OmniControl, SMooDi) are openly cited architectural inspirations and are not used as unverified authorities; there is no uniqueness theorem or self-citation chain forcing the method. The larger concern about synthetic projections substituting for real storyboards is a correctness/robustness limitation, not circularity, and the paper partly acknowledges it in Sec. 7.4. The one genuine circular component is the inference-guidance objective: the reported Avg. Err.-2D is the very function minimized by Eq. 16, so that specific quantitative claim is partially constructed rather than predicted. Hence a moderate partial-circularity score of 4 is appropriate.

Assumptions & free parameters 7 free parameters · 6 assumptions · 0 invented entities

The central claim rests on the quality of HumanML3D as ground truth, on Sketch2Pose for joint detection, and on the assumption that synthetic 2D projections with augmentation represent real storyboards. Loss weights and guidance scales are hand-set to 1 or 7.5, and inference guidance directly optimizes the reported trajectory error.

free parameters (7)

classifier-free guidance scale w_c = 7.5
Hand-set in Sec. 7 Implementation Details; controls the strength of conditioning in Eq. 14.
inference guidance strength tau2 = 1 (by "all hyperparameters set to 1")
Controls L-BFGS trajectory guidance in Eq. 15; directly influences reported Avg. Err.-2D.
loss weights lambda_tr, lambda_key, lambda_r, lambda_c = 1
Set to 1 for training losses in Sec. 4.3 and Sec. 5.
inference guidance iterations K = 4
Number of L-BFGS updates per denoising step (Algo. 1, Sec. 6).
camera augmentation ranges = pitch [0,30], yaw [-45,45], scale [0.8,1.2]
Hand-chosen in supplemental B to mimic storyboard viewpoints; shapes the synthetic test distribution.
joint perturbation noise std = 0.02
Gaussian noise added to synthetic 2D joints to mimic rough strokes (supplemental B).
body proportion scale range = [0.6, 1.6]
Random scaling of body parts during augmentation to simulate disproportionate sketches (supplemental B).

assumptions (6)

domain assumption HumanML3D motion-text pairs are reliable ground truth for training and evaluation.
All quantitative benchmarks derive from this dataset; no independent validation of motion quality or label accuracy is provided.
domain assumption A single action word plus a keypose plus trajectory strokes uniquely specify the intended motion clip.
The system assumes these three conditions resolve ambiguity in the storyboard frame (Sec. 3); real storyboards may rely on context, scene, or sequence.
domain assumption The first verb in the text description is the action word.
Used in supplemental B to construct condition labels; may fail for compound or non-verb descriptions.
domain assumption TMR similarity selects a representative keypose for the action.
The keypose used in training and evaluation is the segment with highest TMR similarity; an incorrect keypose would miscondition the model.
domain assumption Sketch2Pose reliably detects 2D joints in user sketches.
Inference depends on this external tool; the paper notes it can fail on distorted drawings (Sec. 7.4).
ad hoc to paper Orthographic projection with camera and proportion augmentation simulates real storyboard drawings.
The synthetic dataset projects HumanML3D orthographically; real user sketches may have perspective, multiple characters, and varied styles that are not modeled.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Sketch2Anim: Towards Transferring Sketch Storyboards into 3D Animation." pith.science (2026). https://pith.science/paper/B2GJLH5Z

@misc{pith2026250419189,
  author       = {Pith},
  title        = {Pith review of: Sketch2Anim: Towards Transferring Sketch Storyboards into 3D Animation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B2GJLH5Z}},
  note         = {Machine review of arXiv:2504.19189}
}

read the original abstract

Storyboarding is widely used for creating 3D animations. Animators use the 2D sketches in storyboards as references to craft the desired 3D animations through a trial-and-error process. The traditional approach requires exceptional expertise and is both labor-intensive and time-consuming. Consequently, there is a high demand for automated methods that can directly translate 2D storyboard sketches into 3D animations. This task is under-explored to date and inspired by the significant advancements of motion diffusion models, we propose to address it from the perspective of conditional motion synthesis. We thus present Sketch2Anim, composed of two key modules for sketch constraint understanding and motion generation. Specifically, due to the large domain gap between the 2D sketch and 3D motion, instead of directly conditioning on 2D inputs, we design a 3D conditional motion generator that simultaneously leverages 3D keyposes, joint trajectories, and action words, to achieve precise and fine-grained motion control. Then, we invent a neural mapper dedicated to aligning user-provided 2D sketches with their corresponding 3D keyposes and trajectories in a shared embedding space, enabling, for the first time, direct 2D control of motion generation. Our approach successfully transfers storyboards into high-quality 3D motions and inherently supports direct 3D animation editing, thanks to the flexibility of our multi-conditional motion generator. Comprehensive experiments and evaluations, and a user perceptual study demonstrate the effectiveness of our approach.

Figures

Figures reproduced from arXiv: 2504.19189 by the authors.

**Figure 1.** Storyboard animation. Given a storyboard (top row) composed of a few sketch keyposes, joint trajectories, and the corresponding action words, we propose Sketch2Anim, a novel approach that automatically transfers each sketch frame into a 3D motion clip, where the keypose is highlighted in red, and the trajectory is highlighted in green. All these clips are then blended into a complete and coherent animation (bottom r… view at source ↗

**Figure 2.** Traditional 3D animation workflow. Given one frame of a sketch storyboard (top-left), animators first imagine the desired keypose sequence guided by the trajectory and the action word. Two such representative keyposes are shown in blue. These keyposes (including the first keypose) are imported into 3D software (e.g., Blender) as keypose and motion references. The animators then manually place the pre-defined 3D join… view at source ↗

**Figure 3.** Overview of Sketch2Anim. Our pipeline consists of two core modules - the multi-conditional motion generator (Sec. 4) and the 2D-3D neural mapper (Sec. 5). Instead of directly lifting the 2D keypose and trajectory into their 3D counterparts, we train a neural mapper dedicated to aligning the two domains in the embedding space. Because of this shared embedding, it enables the employment of more informative and precise… view at source ↗

Figures from the paper (10 more)

**Figure 4.** Figure 4: User interface and joint detection. (a) In the interface, the white canvas is the main drawing area. Users can sketch or type the action word. We use different pens for character strokes and trajectories. (b) We use Sketch2Pose to detect the 2D joint points (red points…

**Figure 5.** Figure 5: The structure of our motion generator. A trajectory ControlNet is exploited to incorporate the trajectory control, while a keypose adapter is placed between the ControlNet and the diffusion model to inject the keypose condition. The same color and style of the common e…

**Figure 6.** Figure 6: Result Gallery. Three more storyboards are successfully transferred into their high-quality animations. Only the motion clips before blending are shown for clear visualization of keyposes, trajectories, and motion. The complete animation after blending can be seen in t…

**Figure 7.** Figure 7: Synthetic data and its augmentation. Three examples from our dataset are shown in the three rows. As introduced in Sec. 3, our input is the 2D joint points (i.e., red dots), and we draw the stick figure for better visualization. Column (a) demonstrates input data with …

**Figure 8.** Figure 8: Visual comparison. Given two frames of the storyboard (i.e., “Kick” and “Punch”), corresponding results from competitors and ours are shown. Our results faithfully adhere to each frame, while others either have inaccurate joint trajectories or deviated keyposes, leadin…

**Figure 9.** Figure 9: Visual results of the ablation study. Given the conditional 3D keypose (i.e., the purple character) and trajectory (i.e., the yellow curve), resulting motions from alternative methods and ours are shown. Note that the input 3D trajectory is overlaid with the produced m…

**Figure 10.** Figure 10: In total, we have invited 58 participants, and the percentage of times our approach is preferred over the competing methods is displayed in [PITH_FULL_IMAGE:figures/full_fig_p011_10.png]

**Figure 11.** Figure 11: 3D editing. Given one frame of a real-world storyboard (a), we preprocess it to obtain the keypose (i.e., 2D joint points in red) and the trajectory (i.e., curve points in cyan) (b). Our Sketch2Anim produces the animation conditioned on the action word, 2D keypose, a…

**Figure 13.** Figure 13: Impact of conditions. From left to right, we gradually add or change one condition for our method when generating motions at inference time. We highlight the change with the pink color. (a) Only the action word “Walk” is input as the condition. (b) The action word and…

**Figure 14.** Figure 14: Limitations. (a) Our method does not consider character-object interaction, thus the two hands do not hold the golf club at the ending keypose. (b) Without the physical constraints, even if the foot trajectories are correct, the body folds and floats in the air at the…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

10 extracted references · 5 canonical work pages

[2]

In SIGGRAPH Asia 2024 Conference Papers

MotionFix: Text-driven 3d human motion editing. In SIGGRAPH Asia 2024 Conference Papers. German Barquero, Sergio Escalera, and Cristina Palmero. 2024. Seamless human motion composition with blended positional encodings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Darwin Bautista and Rowel Atienza. 2022. Scene tex...

work page 2016
[3]

arXiv preprint arXiv:2407.21136 (2024)

MotionCraft: Crafting Whole-Body Motion with Plug-and-Play Multimodal Controls. arXiv preprint arXiv:2407.21136 (2024). Kirill Brodt and Mikhail Bessmeltsev. 2022. Sketch2Pose: estimating a 3D character pose from a bitmap sketch. ACM Transactions on Graphics (TOG) (2022). Michael Büttner and Simon Clavet. 2015. Motion Matching - The Road to Next Gen Anima...

arXiv 2024
[4]

In ACM SIGGRAPH 2024 Conference Papers

Flexible motion in-betweening with diffusion models. In ACM SIGGRAPH 2024 Conference Papers. Wenxun Dai, Ling-Hao Chen, Jingbo Wang, Jinpeng Liu, Bo Dai, and Yansong Tang

work page 2024
[9]

ACM Transactions on Graphics (2023)

A method for animating children’s drawings of the human figure. ACM Transactions on Graphics (2023). Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020). Jiaming Song, Qinsheng Zhang, Hongxu Yin, Morteza Mardani, Ming-Yu Liu, Jan Kautz, Yongxin Chen, and Arash Vahdat. 2023. Loss-...

arXiv 2023
[10]

A person walks

Motionbert: A unified perspective on learning human motion representations. In Proceedings of the IEEE/CVF International Conference on Computer Vision . ACM Trans. Graph., Vol. 44, No. 4, Article . Publication date: August 2025. 16 • Lei Zhong, Chuan Guo, Yiming Xie, Jiawei Wang, and Changjian Li Supplemental Material In this supplementary material, we pr...

work page 2025
[2006]

In Acm siggraph 2006 courses

A sketching interface for articulated figure animation. In Acm siggraph 2006 courses. 15–es. Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems (2021). Christian Diller and Angela Dai. 2024. Cg-hoi: Contact-guided 3d human-object inter- action generation. In Procee...

arXiv 2021
[2022]

Zhe Li, Weihao Yuan, Yisheng He, Lingteng Qiu, Shenhao Zhu, Xiaodong Gu, Weichao Shen, Yuan Dong, Zilong Dong, and Laurence T

Ganimator: Neural motion synthesis from a single sequence.ACM Transactions on Graphics (TOG) (2022). Zhe Li, Weihao Yuan, Yisheng He, Lingteng Qiu, Shenhao Zhu, Xiaodong Gu, Weichao Shen, Yuan Dong, Zilong Dong, and Laurence T. Yang. 2024b. LaMP: Language- Motion Pretraining for Motion Generation, Retrieval, and Captioning. In arXiv 2410.07093. Juncong Li...

arXiv 2022
[2023]

In Proceedings of the IEEE/CVF International Conference on Computer Vision

Guided Motion Diffusion for Controllable Human Motion Synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision . Lucas Kovar, Michael Gleicher, and Frédéric Pighin. 2023. Motion graphs. In Seminal Graphics Papers: Pushing the Boundaries, Volume 2 . John Lasseter. 1998. Principles of traditional animation applied to 3D computer...

arXiv 2024

Show all 10 references

[2024]

Nikos Athanasiou, Alpár Cseke, Markos Diomataris, Michael J Black, and Gül Varol

SKEL-Betweener: a Neural Motion Rig for Interactive Motion Authoring.ACM Transactions on Graphics (TOG) (2024). Nikos Athanasiou, Alpár Cseke, Markos Diomataris, Michael J Black, and Gül Varol

2024
[2025]

In European Conference on Computer Vision

Motionlcm: Real-time controllable motion generation via latent consistency model. In European Conference on Computer Vision . Springer, 390–408. James Davis, Maneesh Agrawala, Erika Chuang, Zoran Popović, and David Salesin

Pith tools

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

[1] [2]

In SIGGRAPH Asia 2024 Conference Papers

MotionFix: Text-driven 3d human motion editing. In SIGGRAPH Asia 2024 Conference Papers. German Barquero, Sergio Escalera, and Cristina Palmero. 2024. Seamless human motion composition with blended positional encodings. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition . Darwin Bautista and Rowel Atienza. 2022. Scene tex...

work page 2016

[2] [3]

arXiv preprint arXiv:2407.21136 (2024)

MotionCraft: Crafting Whole-Body Motion with Plug-and-Play Multimodal Controls. arXiv preprint arXiv:2407.21136 (2024). Kirill Brodt and Mikhail Bessmeltsev. 2022. Sketch2Pose: estimating a 3D character pose from a bitmap sketch. ACM Transactions on Graphics (TOG) (2022). Michael Büttner and Simon Clavet. 2015. Motion Matching - The Road to Next Gen Anima...

arXiv 2024

[3] [4]

In ACM SIGGRAPH 2024 Conference Papers

Flexible motion in-betweening with diffusion models. In ACM SIGGRAPH 2024 Conference Papers. Wenxun Dai, Ling-Hao Chen, Jingbo Wang, Jinpeng Liu, Bo Dai, and Yansong Tang

work page 2024

[4] [9]

ACM Transactions on Graphics (2023)

A method for animating children’s drawings of the human figure. ACM Transactions on Graphics (2023). Jiaming Song, Chenlin Meng, and Stefano Ermon. 2020. Denoising diffusion implicit models. arXiv preprint arXiv:2010.02502 (2020). Jiaming Song, Qinsheng Zhang, Hongxu Yin, Morteza Mardani, Ming-Yu Liu, Jan Kautz, Yongxin Chen, and Arash Vahdat. 2023. Loss-...

arXiv 2023

[5] [10]

A person walks

Motionbert: A unified perspective on learning human motion representations. In Proceedings of the IEEE/CVF International Conference on Computer Vision . ACM Trans. Graph., Vol. 44, No. 4, Article . Publication date: August 2025. 16 • Lei Zhong, Chuan Guo, Yiming Xie, Jiawei Wang, and Changjian Li Supplemental Material In this supplementary material, we pr...

work page 2025

[6] [2006]

In Acm siggraph 2006 courses

A sketching interface for articulated figure animation. In Acm siggraph 2006 courses. 15–es. Prafulla Dhariwal and Alexander Nichol. 2021. Diffusion models beat gans on image synthesis. Advances in neural information processing systems (2021). Christian Diller and Angela Dai. 2024. Cg-hoi: Contact-guided 3d human-object inter- action generation. In Procee...

arXiv 2021

[7] [2022]

Zhe Li, Weihao Yuan, Yisheng He, Lingteng Qiu, Shenhao Zhu, Xiaodong Gu, Weichao Shen, Yuan Dong, Zilong Dong, and Laurence T

Ganimator: Neural motion synthesis from a single sequence.ACM Transactions on Graphics (TOG) (2022). Zhe Li, Weihao Yuan, Yisheng He, Lingteng Qiu, Shenhao Zhu, Xiaodong Gu, Weichao Shen, Yuan Dong, Zilong Dong, and Laurence T. Yang. 2024b. LaMP: Language- Motion Pretraining for Motion Generation, Retrieval, and Captioning. In arXiv 2410.07093. Juncong Li...

arXiv 2022

[8] [2023]

In Proceedings of the IEEE/CVF International Conference on Computer Vision

Guided Motion Diffusion for Controllable Human Motion Synthesis. In Proceedings of the IEEE/CVF International Conference on Computer Vision . Lucas Kovar, Michael Gleicher, and Frédéric Pighin. 2023. Motion graphs. In Seminal Graphics Papers: Pushing the Boundaries, Volume 2 . John Lasseter. 1998. Principles of traditional animation applied to 3D computer...

arXiv 2024