Pith. sign in

REVIEW 4 major objections 6 minor 30 references

Bolt: Clothing Virtual Characters at Scale

T0 review · 4 major / 6 minor · reviewed 2026-08-16 · deepseek-v4-flash

Pith's one-line read Bolt is an automatic pipeline that refits a source outfit onto many new characters.

desk verdict A plausible industrial pipeline with an honest limitations section, but Equation 5 as written makes the pattern-resizing step impossible, so the paper needs a corrected equation and stronger validation before it is publishable. read the letter →

arxiv 2504.17614 v1 pith:5GQKJLFM submitted 2025-04-24 cs.GR

classification cs.GR
keywords clothsimulationgarmenttransfersewingpatternoptimizationprogressivedrapingskinningweightsigneddistancefieldsoutfitretargetingvirtualhumans
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

Bolt is an automatic pipeline that takes an outfit—3D garment meshes plus their 2D sewing patterns—already fitted to one virtual character and refits it to many other characters. The paper claims this removes the manual tailoring bottleneck that makes large-scale clothed-character production impractical. The pipeline runs in three stages: it transfers the garment geometry and adjusts the sewing pattern to the target body, simulates each garment layer one at a time to untangle and drape the outfit, and transfers the body's skinning weights to the finished cloth. If the claim holds, game studios, animation houses, and synthetic-data generators can populate scenes with thousands of differently shaped clothed characters without an artist touching each outfit.

What carries the argument

The argument rides on four coupled objects. The 3D transfer solves a Stokes-inspired equation system in a narrow band around the body: an inertialess, weakly compressible viscous flow with a unilateral compression penalty, enforced weakly on a regular grid via a compliant penalty, producing the displacement field that moves each cloth vertex. The sewing-pattern optimization transfers warp/weft tangent frames from the source strain state, replaces the target stretch with the source stretch ($F_{\mathrm{target}} = R_{\mathrm{target}} S_{\mathrm{ref}}$), and minimizes a quadratic orthonormality energy together with edge-preserving seam terms using ADMM. The progressive draper simulates one garment layer at a time against a union signed-distance field built from the body and all lower layers, using anisotropic stretch and bending energies plus impulse-based collisions. The rig transfer finds the best body surface point for each cloth vertex by shooting along the cloth normal, then propagates skinning weights with seam-locking and a proximal fallback. Each stage consumes the output of the previous one, and the end-to-end chain is what makes the automation claim stand.

What would settle it

Run the pipeline on a target body generated from a different parametric model or scan, so no per-vertex source-to-target correspondence exists; if the system still produces a draped, rigged outfit, the topology assumption is not load-bearing.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that a multistage offline pipeline can fully automate the refitting of a source outfit onto any target body within one parametric character family: garment 3D positions are displaced by a viscous-flow solve around the body, the 2D sewing pattern is resized by a tangent-frame optimization that preserves seam structure and original strain, each garment layer is successively simulated against the union of the body and previously draped layers to untangle collisions, and skinning weights are transferred from body to cloth using garment normals. The result is a simulation-ready, animation-ready outfit placed on the new character without any manual adjustment. The authors demonstrate the scale this enables: 220 characters generated in under 30 minutes on 64 GPUs, and close to 1000 generated characters shown in the accompanying material.

Load-bearing premise

The transfer step assumes that source and target characters have the same mesh topology and a per-vertex correspondence, so every target vertex has a source vertex to copy displacement from; a target body from any other model family cannot enter the pipeline at all.

Editorial extensions

If this is right

  • A single authored outfit becomes reusable across an entire parametric character population, turning the outfit-fitting problem from a per-character manual task into a batch operation.
  • Multi-layer outfits (demonstrated up to four layers) can be assembled without a global untangling solve, because sequential per-layer simulation against frozen lower layers resolves interpenetration.
  • The output is immediately usable: because skinning weights travel with the garment and are locked across seams, the rigged cloth can be animated without further artist work.
  • At a measured average of roughly six minutes per character on cloud GPUs, generating thousands of varied synthetic humans for training data becomes practical.
  • The system's cost structure is predictable: transfer time depends on garment count and solver convergence, while draping time is the sum of per-layer simulation times, so scaling outfits means scaling simulation, not manual labor.

Reading between the lines

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

  • If the per-vertex correspondence assumption were replaced by a learned or geometric spatial map between arbitrary characters, the same three-stage design would likely transfer outfits between bodies from different generators, scan meshes, or stylized models—the paper's own limitation section points to this as the missing piece.
  • The strain-preserving tangent-frame swap suggests a direct experiment: run the pattern optimizer with and without the $S_{\mathrm{ref}}$ swap and compare the draped silhouettes; the difference would quantify how much preserving original tight-fit strain, rather than simple resizing, is what keeps the garment's style intact.
  • The single-direction layer ordering forbids looks where a later layer must be tucked under an earlier one, such as a sweater over a dress-shirt collar. A simultaneous or sub-layered simulation that allows interleaving would unlock those configurations.
  • Because the simulator forbids seam-group rotations, the system is safe for static drapes and training-data generation, not for full animation; loosening that constraint at runtime would let the same pipeline feed live clothing.
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 / 6 minor

Summary. The paper presents Bolt, an automated pipeline for transferring a manually authored 3D garment outfit from a source virtual character to many target characters. The pipeline has three stages: (1) garment transfer, which computes a 3D displacement field via a Stokes-like flow around the body and then optimizes the 2D sewing pattern to fit the target body; (2) progressive draping, which sequentially simulates each garment layer as a cloth with SDF-based collision handling to untangle and drape the outfit; and (3) rig transfer, which propagates skinning weights from the body to the garments using normal-based correspondence. The authors claim the process requires no human intervention and is feasible for clothing characters at scale, demonstrating results on a variety of body shapes and outfits and reporting wall-clock timings for a 220-character experiment.

Significance. If the system performs as claimed, it addresses a real and combinatorially large problem in digital content creation: refitting multi-garment outfits to many body shapes without manual work. The integration of existing techniques (Stokes transfer, ADMM pattern optimization, XPBD-style simulation, winding-number SDFs) is pragmatic and the scale of the demonstration (hundreds of generated characters) is practically relevant. However, the central contributions are only qualitatively validated, and the written pattern-optimization energy contains a plausibly fatal scaling issue (see major comment 1). The paper also does not release code, so reproducibility hinges on the precision of the algorithm description. The disclosed topology limitation (all bodies from one parametric model) further narrows the headline claim of 'wide range of target body shapes.'

major comments (4)
  1. [Section 4.2.3, Eq. (5)] The pin-to-original penalty term in Eq. (5) has weight 1/(2ε) with ε = 10^-8, i.e., a coefficient of 5×10^7. Unless the 2D pattern coordinates are normalized in a way not stated in the paper, any meaningful resizing of the pattern would incur an enormous energy penalty, making the claimed 'resize the sewing pattern' step effectively impossible. For instance, displacing a single vertex by 1 cm contributes roughly 5×10^7 × (0.01)^2 = 5000 to the second term, while the first term is scaled by triangle areas and is typically orders of magnitude smaller. This contradicts the adjusted patterns shown in Figure 3. The authors should either provide the actual weight used in the implementation, specify a coordinate normalization, or explain why the optimizer can still produce the resized layout; as written, the documented algorithm is internally inconsistent at a load-bearing point.
  2. [Section 8] The central claims of automatic, artifact-free clothing at scale are supported only by still images and the authors' assertion of success. No quantitative quality metrics are reported, such as cloth-body interpenetration volume, maximum/average fabric stretch, pattern-to-body correspondence error, or comparisons to existing refitting methods (e.g., Chen et al. 2024, Brouet et al. 2012, de Goes et al. 2020). To substantiate the claim that the pipeline produces correctly fitted outfits with no human intervention, the paper should include such metrics and, where possible, a comparison or ablation study.
  3. [Section 5.2 and Algorithm 1] The progressive draper is described as running '6 frames' of simulation with 90% global velocity damping at each step. A physical drape normally requires many integration steps for the cloth to settle under gravity and collision contacts; six highly damped frames seem unlikely to reach a static equilibrium for the complex outfits shown. The paper should specify the time-stepping scheme, substep count, and convergence criteria, and provide evidence that six frames (or the equivalent simulated time) is sufficient for the garments to drape and untangle as claimed.
  4. [Section 9.1] The paper honestly discloses that all target characters are derived from a single parametric model and thus share the same topology. This restriction substantially narrows the scope of the 'wide range of target body shapes' claim: the method does not apply to bodies from different generators, manual models, or scans. This is not an internal inconsistency, but the abstract and introduction should more prominently state this scope so readers can calibrate the contribution.
minor comments (6)
  1. [Section 4.2.3] The symbol F_t is used for both the tangent frame matrix in the text and the deformation gradient in Eq. (5); please clarify the notation to avoid confusion.
  2. [Figure 3] An overlay or side-by-side comparison with the source pattern scaled to the target would help readers see the actual resizing; the current figure shows only the source and target patterns without a direct visual comparison.
  3. [Section 8] Tables 2 and 3 are not referenced in the body text; please cite them where the corresponding results are discussed.
  4. [Section 5.1] The winding number threshold of 0.25 is stated, but the paper does not explain how this threshold maps to the inside/outside classification used for the SDF; a brief explanation would be helpful.
  5. [Section 7] The phrase 'proximal transfers' is unusual; consider replacing with 'nearest-point transfers' or clarifying the intended meaning.
  6. [Section 1] The phrase 'resize the sewing pattern' in the abstract and introduction is an oversimplification; the pattern is optimized not merely resized, and the precise behavior is defined later in Section 4.2.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Bolt's transfer, draping, and rigging stages are independent optimization/simulation steps, and the self-citations are infrastructure references only.

full rationale

The claimed derivation chain is self-contained. Garment transfer solves a Stokes-type boundary-value problem (Section 4.1) for a 3D displacement field using the body displacement as a boundary condition, then optimizes the 2D sewing pattern against an explicit energy (Eq. 5) with ADMM edge terms (Section 4.2); neither output is redefined as an input, and the target pattern is not a fitted parameter disguised as a prediction. Progressive draping is a separate per-layer simulation (Section 5), and rig transfer is a geometric weight-propagation step (Section 7). The only self-citations, Warp [Macklin 2022] and XPBD [Macklin et al. 2016], are used as infrastructure and background, not as load-bearing evidence for the central claim, so they do not raise circularity. The disclosed per-vertex topology assumption (Section 9.1) and the reviewer's concern about the large pinning weight in Eq. 5 are correctness, validation, or parameterization issues, not instances of an output reducing by construction to its input.

Assumptions & free parameters 11 free parameters · 7 assumptions · 0 invented entities

The system relies on many hand-set numerical parameters and domain assumptions, most notably the same-topology body correspondence and the sufficiency of a 6-frame, heavily damped per-layer simulation. No new physical entities are introduced.

free parameters (11)
  • compliance gamma = 0.01
    Set in Eq. (1) to allow compression and stretching instead of strict incompressibility; hand-chosen.
  • viscosity nu
    Appears in the Stokes-type transfer equation, but no value is reported; required for the solve and must be tuned.
  • pinning regularization epsilon = 1e-8
    Coefficient of the quadratic pinning term in Eq. (5); chosen so the 2D layout stays close to the original pattern.
  • winding number threshold = 0.25
    Inside/outside cutoff for generalized winding numbers in Section 5.1; hand-set to seal open garment holes.
  • SDF expansion epsilon_sdf = 0.2
    Offset added when unioning SDFs, Table 1 and Section 5.1; chosen to slightly expand the collision field.
  • collision force coefficient = 2500000
    Table 1; collision impulse strength for cloth collisions.
  • collision damping coefficient = 25
    Table 1; damping portion of collision impulses.
  • coulomb friction coefficient = 25
    Table 1; friction portion of collision impulses.
  • global velocity damping = 90% per step
    Section 5.2.3; damps velocities each simulation step to stabilize heavily interpenetrated initial states.
  • simulation frame count = 6 frames
    Section 8; every garment is simulated for only 6 frames; no convergence criterion is reported.
  • garment layer numbers and seam offsets = user or artist specified
    Layer ordering and seam offsets are inputs to the progressive draper; they determine collision ordering and attachment.
assumptions (7)
  • domain assumption Source and target bodies share identical topology with a per-vertex correspondence.
    Section 4.1 defines u_body pointwise on the body surface; Section 9.1 admits all characters come from a single parametric model, limiting generality.
  • domain assumption A 6-frame, 90%-damped per-layer simulation is sufficient to drape and untangle garments.
    Section 8 simulates each garment for 6 frames; no convergence or quality metric is provided.
  • domain assumption Sequential layer-by-layer simulation with frozen lower layers resolves multi-layer interpenetration.
    Section 5 and Algorithm 1 assume later layers see earlier layers as static colliders; Section 9.1 notes no back-propagation is possible.
  • standard math Generalized winding numbers classify inside/outside on open garment meshes at threshold 0.25.
    Section 5.1 relies on Jacobson et al. 2013; the threshold is a hand-set parameter.
  • ad hoc to paper Polar decomposition of triangle tangent frames and swapping stretch, S_ref for S_target, preserves the source fit.
    Section 4.2.2; the stretch-swap is a design heuristic without validation.
  • domain assumption Panels are semantically labeled so proxy mesh generation can filter by type; removed panels have nearby proxy geometry and no significant rotation.
    Section 6 lists these assumptions explicitly.
  • domain assumption Cloth material parameters, including stretch and bending stiffnesses, are known inputs per panel.
    Section 5.2.2 relies on anisotropic constitutive models and industry material parameters.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bolt: Clothing Virtual Characters at Scale." pith.science (2026). https://pith.science/paper/5GQKJLFM

@misc{pith2026250417614,
  author       = {Pith},
  title        = {Pith review of: Bolt: Clothing Virtual Characters at Scale},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5GQKJLFM}},
  note         = {Machine review of arXiv:2504.17614}
}
read the original abstract

Clothing virtual characters is a time-consuming and often manual process. Outfits can be composed of multiple garments, and each garment must be fitted to the unique shape of a character. Since characters can vary widely in size and shape, fitting outfits to many characters is a combinatorially large problem. We present Bolt, a system designed to take outfits originally authored on a source body and fit them to new body shapes via a three stage transfer, drape, and rig process. First, our new garment transfer method transforms each garment's 3D mesh positions to the new character, then optimizes the garment's 2D sewing pattern while maintaining key features of the original seams and boundaries. Second, our system simulates the transferred garments to progressively drape and untangle each garment in the outfit. Finally, the garments are rigged to the new character. This entire process is automatic, making it feasible to clothe characters at scale with no human intervention. Clothed characters are then ready for immediate use in applications such as gaming, animation, synthetic generation, and more.

Figures

Figures reproduced from arXiv: 2504.17614 by the authors.

Figure 1
Figure 1. We present Bolt, an automated method for assembling, refitting, draping, and rigging outfits from a source character to many new characters. On the [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Overview of the Bolt pipeline using an example character. In this example, we assemble a 3-layer outfit from a source character’s body and garments, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. The original sewing pattern adjusted for the size of the target char [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: The blouse originally fitted to the source character, and transferred [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: The progressive draper iteratively simulates each layer to untangle [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 7
Figure 7. Figure 7: Cross-section of garment layers and their associated body surface [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 6
Figure 6. Figure 6: Initial garment (left) and generated proxy mesh (right). [PITH_FULL_IMAGE:figures/full_fig_p008_6.png]
Figure 9
Figure 9. Figure 9: Box plot of transfer times for different outfits fit to a single obese [PITH_FULL_IMAGE:figures/full_fig_p009_9.png]
Figure 12
Figure 12. Figure 12: Multiple outfits on the same character. In this visualization each [PITH_FULL_IMAGE:figures/full_fig_p009_12.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 24 canonical work pages

  1. [8]

    In ACM SIGGRAPH 2024 Conference Papers

    ContourCraft: Learning to Resolve Intersections in Neural Multi-Garment Simulations. In ACM SIGGRAPH 2024 Conference Papers . 1–10. Kai He, Kaixin Yao, Qixuan Zhang, Jingyi Yu, Lingjie Liu, and Lan Xu

  2. [9]

    ACM Transactions on Graphics (TOG) 43, 4 (2024), 1–13

    DressCode: Autoregressively Sewing and Generating Garments from Text Guidance. ACM Transactions on Graphics (TOG) 43, 4 (2024), 1–13. Alec Jacobson, Ladislav Kavan, and Olga Sorkine

  3. [14]

    Yu Lee and Moon-Ryul Jung

    ClothCombo: Modeling Inter-Cloth Interaction for Draping Multi-Layered Clothes.ACM Transactions on Graphics (TOG) 42, 6 (2023), 1–13. Yu Lee and Moon-Ryul Jung

  4. [17]

    https://support.marvelousdesigner.com/hc/en-us/articles/360037396031- FABRIC-PHYSICAL-PROPERTIES-Adjust-Stretch-Weft-Warp-Shear

    FABRIC PHYSICAL PROPERTIES: Adjust Stretch-Weft/Warp, Shear. https://support.marvelousdesigner.com/hc/en-us/articles/360037396031- FABRIC-PHYSICAL-PROPERTIES-Adjust-Stretch-Weft-Warp-Shear. Accessed: 2023-11-27. Shanthika Naik, Kunwar Singh, Astitva Srivastava, Dhawal Sirikonda, Amit Raj, Varun Jampani, and Avinash Sharma

  5. [18]

    arXiv preprint arXiv:2401.03108 (2024)

    Dress-Me-Up: A Dataset & Method for Self- Supervised 3D Garment Retargeting. arXiv preprint arXiv:2401.03108 (2024). Matthew Overby, George E. Brown, Jie Li, and Rahul Narain

  6. [19]

    IEEE Trans

    ADMM Projective Dynamics: Fast Simulation of Hyperelastic Models with Dynamic Constraints. IEEE Trans. Vis. Comput. Graph. 23, 10 (2017), 2222–2234. http://dblp.uni-trier.de/db/ journals/tvcg/tvcg23.html#OverbyBLN17 Nico Pietroni, Corentin Dumery, Raphael Guenot-Falque, Mark Liu, Teresa Vidal-Calleja, and Olga Sorkine-Hornung

  7. [20]

    Computational Pattern Making from 3D Garment Models

    Computational Pattern Making from 3D Garment Models. arXiv:2202.10272 [cs.GR] Gerard Pons-Moll, Sergi Pujades, Sonny Hu, and Michael J Black

  8. [21]

    ACM Transactions on Graphics (ToG) 36, 4 (2017), 1–15

    ClothCap: Seamless 4D clothing capture and retargeting. ACM Transactions on Graphics (ToG) 36, 4 (2017), 1–15. Xavier Provot

Show all 30 references
  1. [28]

    ACM Transactions on Graphics (TOG) 37, 4 (2018), 1–13

    Rule-free sewing pattern adjustment with precision and efficiency. ACM Transactions on Graphics (TOG) 37, 4 (2018), 1–13. Huamin Wang and Yin Yang

  2. [1987]

    In Proceedings of the 14th Annual Conference on Computer Graphics and Interactive Techniques (SIGGRAPH ’87)

    Elastically de- formable models. In Proceedings of the 14th Annual Conference on Computer Graphics and Interactive Techniques (SIGGRAPH ’87) . Association for Computing Machinery, New York, NY, USA, 205–214. https://doi.org/10.1145/37401.37427 Pascal Volino and Nadia Magnenat-Thalmann

  3. [1997]

    In Computer Animation and Simulation’97: Proceedings of the Eurographics Workshop in Budapest, Hungary, September 2–3, 1997

    Collision and self-collision handling in cloth model dedicated to design garments. In Computer Animation and Simulation’97: Proceedings of the Eurographics Workshop in Budapest, Hungary, September 2–3, 1997 . Springer, 177–

  4. [2002]

    ACM Trans

    Robust treatment of collisions, contact and friction for cloth animation. ACM Trans. Graph. 21, 3 (2002). Remi Brouet, Alla Sheffer, Laurence Boissieux, and Marie-Paule Cani

  5. [2003]

    ACM Trans- actions on Graphics (TOG) 22, 3 (2003), 862–870

    Untangling cloth. ACM Trans- actions on Graphics (TOG) 22, 3 (2003), 862–870. Sofien Bouaziz, Sebastian Martin, Tiantian Liu, Ladislav Kavan, and Mark Pauly

  6. [2005]

    In ACM SIGGRAPH 2005 Sketches

    Cloth simulation on the gpu. In ACM SIGGRAPH 2005 Sketches . 39–es. 12 • Leaf et al. Table

  7. [2006]

    ACM Transactions on Graphics (TOG) 25, 3 (2006), 1154–1159

    Resolving surface collisions through intersection contour minimization. ACM Transactions on Graphics (TOG) 25, 3 (2006), 1154–1159. Huamin Wang

  8. [2007]

    In Geome- try Processing, Alexander Belyaev and Michael Garland (Eds.)

    As-Rigid-As-Possible Surface Modeling. In Geome- try Processing, Alexander Belyaev and Michael Garland (Eds.). The Eurographics Association. https://doi.org/10.2312/SGP/SGP07/109-116 Min Tang, Tongtong Wang, Zhongyuan Liu, Ruofeng Tong, and Dinesh Manocha

  9. [2011]

    In ACM SIGGRAPH 2011 papers

    Example-based elastic materials. In ACM SIGGRAPH 2011 papers. 1–8. Marvelous Designer

  10. [2012]

    ACM Transactions on Graphics 31, 4 (2012), Article–No

    Design preserving garment transfer. ACM Transactions on Graphics 31, 4 (2012), Article–No. Thomas Buffet, Damien Rohmer, Loic Barthe, Laurence Boissieux, and Marie-Paule Cani

  11. [2013]

    ACM Trans

    Robust Inside-Outside Segmen- tation using Generalized Winding Numbers. ACM Trans. Graph. 32, 4 (2013). Theodore Kim

  12. [2014]

    ACM Trans- actions on Graphics (TOG) 33, 4 (2014), 1–11

    Projective dynamics: Fusing constraint projections for fast simulation. ACM Trans- actions on Graphics (TOG) 33, 4 (2014), 1–11. Robert Bridson, Ronald Fedkiw, and John Anderson

  13. [2015]

    In Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation (Copenhagen, Denmark) (SCA ’14)

    Optimization integrator for large time steps. In Proceedings of the ACM SIGGRAPH/Eurographics Symposium on Computer Animation (Copenhagen, Denmark) (SCA ’14). Eurographics Association, Goslar, DEU, 31–40. Yotam Gingold, Adrian Secord, Jefferson Y. Han, and Eitan Grinspun. 2004...

  14. [2016]

    ACM Transactions on Graphics (TOG) 35, 6 (2016), 1–10

    Descent methods for elastic body simulation on the GPU. ACM Transactions on Graphics (TOG) 35, 6 (2016), 1–10. Martin Wicke, Hermes Lanker, and Markus Gross

  15. [2017]

    Acm Transactions on Graphics (TOG) 36, 3 (2017), 1–16

    Quasi-newton methods for real-time simulation of hyperelastic materials. Acm Transactions on Graphics (TOG) 36, 3 (2017), 1–16. Miles Macklin

  16. [2018]

    ACM Transaction on Graphics (Proceedings of SIGGRAPH Asia) 37, 6 (November 2018), 204:1–10

    I-Cloth: Incremental Collision Handling for GPU-Based Interactive Cloth Simulation. ACM Transaction on Graphics (Proceedings of SIGGRAPH Asia) 37, 6 (November 2018), 204:1–10. Demetri Terzopoulos, John Platt, Alan Barr, and Kurt Fleischer

  17. [2019]

    ACM Transactions on Graphics (TOG) 38, 4 (2019), 1–12

    Implicit untangling: A robust solution for modeling layered clothing. ACM Transactions on Graphics (TOG) 38, 4 (2019), 1–12. Hsiao-yu Chen, Egor Larionov, Ladislav Kavan, Gene Lin, Doug Roble, Olga Sorkine- Hornung, and Tuur Stuyck

  18. [2020]

    ACM SIGGRAPH / Eurographics Symposium on Computer Animation 39, 8 (2020)

    A Finite Element Formulation of Baraff-Witkin Cloth. ACM SIGGRAPH / Eurographics Symposium on Computer Animation 39, 8 (2020). Maria Korosteleva and Sung-Hee Lee

  19. [2021]

    Graphical Models 115 (2021), 101106

    Learning a shared deformation space for efficient design-preserving garment transfer. Graphical Models 115 (2021), 101106. Olga Sorkine and Marc Alexa

  20. [2022]

    ACM Trans

    NeuralTailor: Reconstructing Sewing Pattern Structures from 3D Point Clouds of Garments. ACM Trans. Graph. 41, 4 (2022), 16 pages. https://doi.org/10.1145/3528223.3530179 Maria Korosteleva and Olga Sorkine-Hornung

  21. [2023]

    ACM Transaction on Graphics 42, 6 (2023), 16 pages

    GarmentCode: Programming Parametric Sewing Patterns. ACM Transaction on Graphics 42, 6 (2023), 16 pages. https://doi.org/10.1145/3618351 SIGGRAPH ASIA 2023 issue. Dohae Lee, Hyun Kang, and In-Kwon Lee

  22. [2024]

    arXiv preprint arXiv:2405.19148 (2024)

    Dress Anyone: Automatic Physically-Based Garment Pattern Refitting. arXiv preprint arXiv:2405.19148 (2024). Fernando de Goes, Donald Fong, and Meredith O’Malley

Pith tools

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