pith. sign in

arxiv: 2605.15824 · v1 · pith:ETMRAALJnew · submitted 2026-05-15 · 💻 cs.CV

FashionChameleon: Towards Real-Time and Interactive Human-Garment Video Customization

Quanjian Song , Yefeng Shen , Mengting Chen , Hao Sun , Jinsong Lan , Xiaoyong Zhu , Bo Zheng , Liujuan Cao This is my paper

Pith reviewed 2026-05-20 20:05 UTC · model grok-4.3

classification 💻 cs.CV
keywords garment video customizationreal-time video generationinteractive editingmotion coherencein-context learningKV cache reschedulingvideo distillationautoregressive models
0
0 comments X

The pith

A framework achieves real-time interactive garment switching in human videos while preserving motion coherence using only single-garment training data.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper seeks to enable users to change garments on the fly during video generation without losing the consistency of the person's movements. Current methods require lots of multi-garment data and run too slow for interactive use. By training a teacher model on mismatched single-garment pairs to learn coherence implicitly, then distilling it and adding a way to reschedule key-value caches on the fly, the approach delivers fast generation. This would matter for online shopping try-ons and quick video content creation where users want to experiment with outfits instantly.

Core claim

The authors show that a combination of in-context learning on mismatched single-garment video pairs for a teacher model, followed by streaming distillation with in-context teacher forcing and gradient-reweighted distribution matching, plus training-free KV cache rescheduling for garment KV refresh, historical KV withdraw, and reference KV disentangle, allows the model to support interactive multi-garment customization in autoregressive video generation while keeping motion coherence and running at real-time speeds of 23.8 FPS on a single GPU, which is 30-180 times faster than baselines.

What carries the argument

Training-Free KV Cache Rescheduling, which refreshes garment-specific keys and values, withdraws historical ones, and disentangles reference information to enable seamless garment switches without retraining.

Load-bearing premise

That deliberately mismatching the reference garment image during training on single-garment videos will teach the model to keep motion coherent when garments are switched at generation time.

What would settle it

Generate a video where the garment is switched halfway and check if the person's limb movements and body posture continue exactly as in the original motion sequence without jumps or inconsistencies.

Figures

Figures reproduced from arXiv: 2605.15824 by Bo Zheng, Hao Sun, Jinsong Lan, Liujuan Cao, Mengting Chen, Quanjian Song, Xiaoyong Zhu, Yefeng Shen.

Figure 1
Figure 1. Figure 1: Given a reference image and a sequence of garment images, FashionChameleon generate [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Average performance (Cur., GME, Amp., Smoo., and VQ) and inference speed com￾parison across different approaches. Despite this progress, existing customization meth￾ods mainly focus on human-centric subject con￾sistency, with comparatively less emphasis on fine-grained human attributes. Among these at￾tributes, garment-level customization is particu￾larly desirable in practical applications such as filmmak… view at source ↗
Figure 3
Figure 3. Figure 3: Overall pipeline of FashionChameleon: Teacher Model with In-Context Learning, Streaming Distillation with In-Context Learning, and Training-Free KV Cache Rescheduling. is generated autoregressively. Self-Forcing [28] further improves this paradigm with self-rolling, conditioning on self-generated rather than ground-truth history to better align training with inference. To avoid training from scratch, most … view at source ↗
Figure 4
Figure 4. Figure 4: (Left) Generated sequences during garment switching. Directly refreshing the garment KV [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparison of our FashionChameleon with other baselines. Due to space [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Additional applications of FashionChameleon. It supports both [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Qualitative ablation of Gradient-Reweighted Distribution Matching Distillation (DMD) and Reference KV Disentangle. Gradient-Reweighted DMD alleviates motion collapse during extrap￾olation, while Reference KV Disentangle further enhances consistency during garment switching [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: The high-quality data curation pipeline of FashionChameleon. It consists of four stages: [PITH_FULL_IMAGE:figures/full_fig_p014_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Data analysis and representative samples of HGC-Bench. (a) A word cloud generated from [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Quantitative results of the human evaluation. We compare our FashionChameleon [PITH_FULL_IMAGE:figures/full_fig_p017_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Additional qualitative comparison between our FashionChameleon and other baselines. [PITH_FULL_IMAGE:figures/full_fig_p019_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Additional qualitative comparison between our FashionChameleon and other baselines. [PITH_FULL_IMAGE:figures/full_fig_p019_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Additional results for short video customization using our FashionChameleon. [PITH_FULL_IMAGE:figures/full_fig_p020_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Additional results for short video customization using our FashionChameleon. [PITH_FULL_IMAGE:figures/full_fig_p020_14.png] view at source ↗
Figure 16
Figure 16. Figure 16: Unlike conventional methods that require a reference image to be specified in advance, [PITH_FULL_IMAGE:figures/full_fig_p020_16.png] view at source ↗
Figure 15
Figure 15. Figure 15: Additional visualizations for interactive multi-garment video customization using our [PITH_FULL_IMAGE:figures/full_fig_p021_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Additional visualizations for interactive multi-garment video customization using our [PITH_FULL_IMAGE:figures/full_fig_p021_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Additional long video extrapolation visualizations of our FashionChameleon. [PITH_FULL_IMAGE:figures/full_fig_p022_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Additional long video extrapolation visualizations of our FashionChameleon. [PITH_FULL_IMAGE:figures/full_fig_p022_18.png] view at source ↗
read the original abstract

Human-centric video customization, particularly at the garment level, has shown significant commercial value. However, existing approaches cannot support low-latency and interactive garment control, which is crucial for applications such as e-commerce and content creation. This paper studies how to achieve interactive multi-garment video customization while preserving motion coherence using only single-garment video data. We present FashionChameleon, a real-time and interactive framework for human-garment customization in autoregressive video generation, where users can interactively switch garment during generation. FashionChameleon consists of three key techniques: (i) Instead of training on multi-garment video data, we train a Teacher Model with In-Context Learning on a single reference-garment pair. By retaining the image-to-video training paradigm while enforcing a mismatch between the reference and garment image, the model is encouraged to implicitly preserve coherence during single-garment switching. (ii) To achieve consistency and efficiency during generation, we introduce Streaming Distillation with In-Context Learning, which fine-tunes the model with in-context teacher forcing and improves extrapolation consistency via gradient-reweighted distribution matching distillation. (iii) To extend the model for interactive multi-garment video customization, we propose Training-Free KV Cache Rescheduling, which includes garment KV refresh, historical KV withdraw, and reference KV disentangle to achieve garment switching while preserving motion coherence. Our FashionChameleon uniquely supports interactive customization and consistent long-video extrapolation, while achieving real-time generation at 23.8 FPS on a single GPU, 30-180$\times$ faster than existing baselines.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 2 minor

Summary. The manuscript presents FashionChameleon, a real-time framework for interactive multi-garment video customization in autoregressive human video generation. Using only single-garment video data, it introduces three techniques: (i) a teacher model trained with in-context learning on deliberately mismatched reference-garment pairs to implicitly encourage motion coherence preservation during switches, (ii) streaming distillation with in-context learning and gradient-reweighted distribution matching for consistency and efficiency, and (iii) training-free KV cache rescheduling (including garment KV refresh, historical KV withdraw, and reference KV disentangle) to enable mid-video garment switching. The work reports 23.8 FPS on a single GPU (30-180× faster than baselines) and claims support for consistent long-video extrapolation.

Significance. If the empirical validation holds, the result would represent a meaningful advance for low-latency interactive applications in e-commerce and content creation. The combination of single-garment-only training, training-free switching mechanism, and real-time performance addresses practical limitations in prior human-centric video generation approaches. The reported speed and the design of the KV cache rescheduling are concrete strengths that could enable new interactive workflows.

major comments (1)
  1. [Abstract, first key technique] Description of first key technique (mismatch training on single-garment pairs): The central claim that deliberately mismatching the reference garment image while training on single-garment video pairs will implicitly teach reliable motion-garment disentanglement and coherence preservation for mid-video switches rests on an indirect inductive bias. No explicit temporal consistency loss, switch-specific supervision, or autoregressive error analysis is described, raising the risk that small inconsistencies accumulate over long horizons or complex motions. This assumption is load-bearing for the interactive customization and long-video extrapolation claims; an ablation (matched vs. mismatched training) with quantitative coherence metrics (e.g., temporal consistency scores or user studies) is required to substantiate it.
minor comments (2)
  1. [Abstract] The speed claim of '30-180× faster' would be clearer if the specific baselines, their hardware configurations, and the exact metric (e.g., latency per frame) were stated explicitly.
  2. [Description of third technique] Notation for 'KV Cache Rescheduling' components (garment KV refresh, historical KV withdraw, reference KV disentangle) could be formalized with a short diagram or pseudocode to improve reproducibility.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the constructive feedback and detailed analysis of our work on FashionChameleon. We address the major comment regarding the mismatch training technique below. We agree that additional evidence would strengthen the presentation and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Abstract, first key technique] Description of first key technique (mismatch training on single-garment pairs): The central claim that deliberately mismatching the reference garment image while training on single-garment video pairs will implicitly teach reliable motion-garment disentanglement and coherence preservation for mid-video switches rests on an indirect inductive bias. No explicit temporal consistency loss, switch-specific supervision, or autoregressive error analysis is described, raising the risk that small inconsistencies accumulate over long horizons or complex motions. This assumption is load-bearing for the interactive customization and long-video extrapolation claims; an ablation (matched vs. mismatched training) with quantitative coherence metrics (e.g., temporal consistency scores or user studies) is required to substantiate it.

    Authors: We appreciate the referee pointing out the indirect nature of this inductive bias. The core motivation is that standard matched reference-garment training allows the model to shortcut by directly copying appearance from the reference image, bypassing the need to preserve motion coherence from the video context. By deliberately mismatching the reference garment while keeping the single-garment video pairs, the in-context learning objective forces the model to attend to the temporal sequence for motion information rather than relying on appearance alignment. This is retained within the standard image-to-video paradigm without adding new losses. Our reported results on long-video extrapolation and interactive switching (including 23.8 FPS performance and coherence in multi-garment scenarios) provide supporting evidence that inconsistencies do not accumulate in practice, aided by the training-free KV cache mechanisms. That said, we agree a controlled ablation (matched vs. mismatched) with metrics such as temporal consistency scores and user studies would make the claim more robust. We will include this ablation in the revised manuscript. revision: yes

Circularity Check

0 steps flagged

Derivation chain is self-contained with no circular reductions

full rationale

The paper proposes FashionChameleon as a framework using three explicit techniques to enable interactive multi-garment video customization from single-garment data only. Technique (i) trains a teacher model on mismatched reference-garment pairs to encourage implicit motion coherence; (ii) applies streaming distillation with in-context learning and gradient-reweighted matching; (iii) uses training-free KV cache rescheduling for switches. These are design choices and training procedures whose outputs are evaluated empirically against baselines (e.g., 23.8 FPS, 30-180× speedup). No equations, fitted parameters renamed as predictions, or load-bearing self-citations appear that would reduce any central claim to its inputs by construction. The derivation therefore remains independent of the target results.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the paper introduces three new procedural components rather than new physical entities or free parameters; no explicit fitted constants or ad-hoc axioms are stated.

axioms (1)
  • domain assumption Autoregressive video generation models can maintain motion coherence when reference and target garment images are deliberately mismatched during training.
    Invoked in the description of the Teacher Model with In-Context Learning.

pith-pipeline@v0.9.0 · 5843 in / 1365 out tokens · 52553 ms · 2026-05-20T20:05:31.792661+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

102 extracted references · 102 canonical work pages · 17 internal anchors

  1. [1]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” in Conference on Neural Information Processing Systems, 2020

    work page 2020

  2. [2]

    Flow Matching for Generative Modeling

    Y . Lipman, R. T. Chen, H. Ben-Hamu, M. Nickel, and M. Le, “Flow matching for generative modeling,”arXiv preprint arXiv:2210.02747, 2022

    work page internal anchor Pith review Pith/arXiv arXiv 2022

  3. [3]

    CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer

    Z. Yang, J. Teng, W. Zheng, M. Ding, S. Huang, J. Xu, Y . Yang, W. Hong, X. Zhang, G. Feng et al., “Cogvideox: Text-to-video diffusion models with an expert transformer,”arXiv preprint arXiv:2408.06072, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  4. [4]

    HunyuanVideo: A Systematic Framework For Large Video Generative Models

    W. Kong, Q. Tian, Z. Zhang, R. Min, Z. Dai, J. Zhou, J. Xiong, X. Li, B. Wu, J. Zhanget al., “Hunyuanvideo: A systematic framework for large video generative models,” arXiv preprint arXiv:2412.03603, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  5. [5]

    Wan: Open and Advanced Large-Scale Video Generative Models

    T. Wan, A. Wang, B. Ai, B. Wen, C. Mao, C.-W. Xie, D. Chen, F. Yu, H. Zhao, J. Yang et al., “Wan: Open and advanced large-scale video generative models,” arXiv preprint arXiv:2503.20314, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  6. [6]

    Realcam-i2v: Real-world image-to-video generation with interactive complex camera control,

    T. Li, G. Zheng, R. Jiang, S. Zhan, T. Wu, Y . Lu, Y . Lin, C. Deng, Y . Xiong, M. Chenet al., “Realcam-i2v: Real-world image-to-video generation with interactive complex camera control,” in International Conference on Computer Vision, 2025

    work page 2025

  7. [7]

    Drivegenvlm: Real-world video generation for vision language model based autonomous driving,

    Y . Fu, A. Jain, X. Chen, Z. Mo, and X. Di, “Drivegenvlm: Real-world video generation for vision language model based autonomous driving,” in IEEE International automated vehicle validation conference, 2024

    work page 2024

  8. [8]

    Worldwander: Bridging egocentric and exocentric worlds in video generation,

    Q. Song, Y . Song, K. Peng, Y . Gao, and M. Z. Shou, “Worldwander: Bridging egocentric and exocentric worlds in video generation,”arXiv preprint arXiv:2511.22098, 2025

    work page arXiv 2025

  9. [9]

    Customvideo: Customizing text-to-video generation with multiple subjects,

    Z. Wang, A. Li, L. Zhu, Y . Guo, Q. Dou, and Z. Li, “Customvideo: Customizing text-to-video generation with multiple subjects,”IEEE Transactions on Multimedia, 2026

    work page 2026

  10. [10]

    Disenstudio: Cus- tomized multi-subject text-to-video generation with disentangled spatial control,

    H. Chen, X. Wang, Y . Zhang, Y . Zhou, Z. Zhang, S. Tang, and W. Zhu, “Disenstudio: Cus- tomized multi-subject text-to-video generation with disentangled spatial control,” in ACM International Conference on Multimedia, 2024

    work page 2024

  11. [11]

    Id- animator: Zero-shot identity-preserving human video gener- ation.arXiv:2404.15275, 2024

    X. He, Q. Liu, S. Qian, X. Wang, T. Hu, K. Cao, K. Yan, and J. Zhang, “Id-animator: Zero-shot identity-preserving human video generation,”arXiv preprint arXiv:2404.15275, 2024

    work page arXiv 2024

  12. [12]

    Identity-preserving text- to-video generation by frequency decomposition,

    S. Yuan, J. Huang, X. He, Y . Ge, Y . Shi, L. Chen, J. Luo, and L. Yuan, “Identity-preserving text- to-video generation by frequency decomposition,” in IEEE Conference on Computer Vision and Pattern Recognition, 2025. 10

    work page 2025

  13. [13]

    Phantom: Subject- consistent video generation via cross-modal alignment

    L. Liu, T. Ma, B. Li, Z. Chen, J. Liu, G. Li, S. Zhou, Q. He, and X. Wu, “Phantom: Subject- consistent video generation via cross-modal alignment,” arXiv preprint arXiv:2502.11079, 2025

    work page arXiv 2025

  14. [14]

    Vace: All-in-one video creation and editing,

    Z. Jiang, Z. Han, C. Mao, J. Zhang, Y . Pan, and Y . Liu, “Vace: All-in-one video creation and editing,” inInternational Conference on Computer Vision, 2025

    work page 2025

  15. [15]

    Stand-in: A lightweight and plug-and-play identity control for video generation,

    B. Xue, Z.-P. Duan, Q. Yan, W. Wang, H. Liu, C.-L. Guo, C. Li, C. Li, and J. Lyu, “Stand-in: A lightweight and plug-and-play identity control for video generation,” inIEEE Conference on Computer Vision and Pattern Recognition, 2026

    work page 2026

  16. [16]

    Scalable diffusion models with transformers,

    W. Peebles and S. Xie, “Scalable diffusion models with transformers,” in International Conference on Computer Vision, 2023

    work page 2023

  17. [17]

    Bindweave: Subject-consistent video generation via cross-modal integration,

    Z. Li, D. Qian, K. Su, Q. Diao, X. Xia, C. Liu, W. Yang, T. Zhang, and Z. Yuan, “Bindweave: Subject-consistent video generation via cross-modal integration,”arXiv preprint arXiv:2510.00438, 2025

    work page arXiv 2025

  18. [18]

    Magref: Masked guidance for any-reference video generation with subject disentanglement

    Y . Deng, Y . Yin, X. Guo, Y . Wang, J. Z. Fang, S. Yuan, Y . Yang, A. Wang, B. Liu, H. Huanget al., “Magref: Masked guidance for any-reference video generation with subject disentanglement,” arXiv preprint arXiv:2505.23742, 2025

    work page arXiv 2025

  19. [19]

    Skyreels-a2: Compose anything in video diffusion transformers.arXiv preprint arXiv:2504.02436, 2025

    Z. Fei, D. Li, D. Qiu, J. Wang, Y . Dou, R. Wang, J. Xu, M. Fan, G. Chen, Y . Liet al., “Skyreels- a2: Compose anything in video diffusion transformers,” arXiv preprint arXiv:2504.02436, 2025

    work page arXiv 2025

  20. [20]

    Kaleido: Open-sourced multi-subject reference video generation model,

    Z. Zhang, J. Teng, Z. Yang, T. Cao, C. Wang, X. Gu, J. Tang, D. Guo, and M. Wang, “Kaleido: Open-sourced multi-subject reference video generation model,” arXiv preprint arXiv:2510.18573, 2025

    work page arXiv 2025

  21. [21]

    Motionctrl: A unified and flexible motion controller for video generation,

    Z. Wang, Z. Yuan, X. Wang, Y . Li, T. Chen, M. Xia, P. Luo, and Y . Shan, “Motionctrl: A unified and flexible motion controller for video generation,” inACM SIGGRAPH Conference on Computer Graphics and Interactive Techniques, 2024

    work page 2024

  22. [22]

    Lightmotion: A light and tuning-free method for simulating camera motion in video generation,

    Q. Song, Z. Lin, Z. Zeng, Z. Zhang, L. Cao, and R. Ji, “Lightmotion: A light and tuning-free method for simulating camera motion in video generation,”arXiv preprint arXiv:2503.06508, 2025

    work page arXiv 2025

  23. [23]

    Artifical intelligence in e-commerce: Applications, implications and challenges,

    H. A. Lari, K. Vaishnava, and K. Manu, “Artifical intelligence in e-commerce: Applications, implications and challenges,”Asian Journal of Management, 2022

    work page 2022

  24. [24]

    Photomaker: Customizing realistic human photos via stacked id embedding,

    Z. Li, M. Cao, X. Wang, Z. Qi, M.-M. Cheng, and Y . Shan, “Photomaker: Customizing realistic human photos via stacked id embedding,” inIEEE Conference on Computer Vision and Pattern Recognition, 2024

    work page 2024

  25. [25]

    Univst: A unified framework for training-free localized video style transfer,

    Q. Song, M. Lin, W. Zhan, S. Yan, L. Cao, and R. Ji, “Univst: A unified framework for training-free localized video style transfer,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025

    work page 2025

  26. [26]

    Objectadd: adding objects into image via a training-free diffusion modification fashion,

    Z. Zhang, M. Lin, Q. Song, Y . Zhang, and R. Ji, “Objectadd: adding objects into image via a training-free diffusion modification fashion,”Pattern Recognition, 2025

    work page 2025

  27. [27]

    From slow bidirectional to fast autoregressive video diffusion models,

    T. Yin, Q. Zhang, R. Zhang, W. T. Freeman, F. Durand, E. Shechtman, and X. Huang, “From slow bidirectional to fast autoregressive video diffusion models,” inIEEE Conference on Computer Vision and Pattern Recognition, 2025

    work page 2025

  28. [28]

    Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion

    X. Huang, Z. Li, G. He, M. Zhou, and E. Shechtman, “Self forcing: Bridging the train-test gap in autoregressive video diffusion,”arXiv preprint arXiv:2506.08009, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  29. [29]

    Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation

    H. Zhu, M. Zhao, G. He, H. Su, C. Li, and J. Zhu, “Causal forcing: Autoregressive diffusion distillation done right for high-quality real-time interactive video generation,”arXiv preprint arXiv:2602.02214, 2026

    work page internal anchor Pith review arXiv 2026

  30. [30]

    Flashvsr: Towards real- time diffusion-based streaming video super-resolution.arXiv preprint arXiv:2510.12747, 2025

    J. Zhuang, S. Guo, X. Cai, X. Li, Y . Liu, C. Yuan, and T. Xue, “Flashvsr: Towards real-time diffusion-based streaming video super-resolution,”arXiv preprint arXiv:2510.12747, 2025

    work page arXiv 2025

  31. [31]

    Live Avatar: Streaming Real-time Audio-Driven Avatar Generation with Infinite Length

    Y . Huang, H. Guo, F. Wu, S. Zhang, S. Huang, Q. Gan, L. Liu, S. Zhao, E. Chen, J. Liuet al., “Live avatar: Streaming real-time audio-driven avatar generation with infinite length,”arXiv preprint arXiv:2512.04677, 2025. 11

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  32. [32]

    Motionstream: Real-time video gen- eration with interactive motion controls.arXiv preprint arXiv:2511.01266,

    J. Shin, Z. Li, R. Zhang, J.-Y . Zhu, J. Park, E. Shechtman, and X. Huang, “Motionstream: Real-time video generation with interactive motion controls,”arXiv preprint arXiv:2511.01266, 2025

    work page arXiv 2025

  33. [33]

    All are worth words: A vit backbone for diffusion models,

    F. Bao, S. Nie, K. Xue, Y . Cao, C. Li, H. Su, and J. Zhu, “All are worth words: A vit backbone for diffusion models,” inIEEE Conference on Computer Vision and Pattern Recognition, 2023

    work page 2023

  34. [34]

    Diffusion forc- ing: Next-token prediction meets full-sequence diffusion,

    B. Chen, D. Martí Monsó, Y . Du, M. Simchowitz, R. Tedrake, and V . Sitzmann, “Diffusion forc- ing: Next-token prediction meets full-sequence diffusion,” inConference on Neural Information Processing Systems, 2024

    work page 2024

  35. [35]

    VideoPoet: A Large Language Model for Zero-Shot Video Generation

    D. Kondratyuk, L. Yu, X. Gu, J. Lezama, J. Huang, G. Schindler, R. Hornung, V . Birodkar, J. Yan, M.-C. Chiu et al., “Videopoet: A large language model for zero-shot video generation,” arXiv preprint arXiv:2312.14125, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  36. [36]

    Autoregressive Model Beats Diffusion: Llama for Scalable Image Generation

    P. Sun, Y . Jiang, S. Chen, S. Zhang, B. Peng, P. Luo, and Z. Yuan, “Autoregressive model beats diffusion: Llama for scalable image generation,” arXiv preprint arXiv:2406.06525, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  37. [37]

    Improved distribution matching distillation for fast image synthesis,

    T. Yin, M. Gharbi, T. Park, R. Zhang, E. Shechtman, F. Durand, and W. T. Freeman, “Improved distribution matching distillation for fast image synthesis,” inConference on Neural Information Processing Systems, 2024

    work page 2024

  38. [38]

    Rolling Forcing: Autoregressive Long Video Diffusion in Real Time

    K. Liu, W. Hu, J. Xu, Y . Shan, and S. Lu, “Rolling forcing: Autoregressive long video diffusion in real time,”arXiv preprint arXiv:2509.25161, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  39. [39]

    Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillation

    Y . Lu, Y . Zeng, H. Li, H. Ouyang, Q. Wang, K. L. Cheng, J. Zhu, H. Cao, Z. Zhang, X. Zhuet al., “Reward forcing: Efficient streaming video generation with rewarded distribution matching distillation,”arXiv preprint arXiv:2512.04678, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  40. [40]

    Infinity-rope: Action-controllable infinite video generation emerges from autoregressive self- rollout.arXiv preprint arXiv:2511.20649, 2025

    H. Yesiltepe, T. H. S. Meral, A. K. Akan, K. Oktay, and P. Yanardag, “Infinity-rope: Action- controllable infinite video generation emerges from autoregressive self-rollout,”arXiv preprint arXiv:2511.20649, 2025

    work page arXiv 2025

  41. [41]

    LongLive: Real-time Interactive Long Video Generation

    S. Yang, W. Huang, R. Chu, Y . Xiao, Y . Zhao, X. Wang, M. Li, E. Xie, Y . Chen, Y . Luet al., “Longlive: Real-time interactive long video generation,” arXiv preprint arXiv:2509.22622, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  42. [42]

    Huang, J

    S. Huang, J. Wu, Q. Zhou, S. Miao, and M. Long, “Vid2world: Crafting video diffusion models to interactive world models,”arXiv preprint arXiv:2505.14357, 2025

    work page arXiv 2025

  43. [43]

    Yume-1.5: A text-controlled interactive world generation model.arXiv preprint arXiv:2512.22096,

    X. Mao, Z. Li, C. Li, X. Xu, K. Ying, T. He, J. Pang, Y . Qiao, and K. Zhang, “Yume-1.5: A text-controlled interactive world generation model,”arXiv preprint arXiv:2512.22096, 2025

    work page arXiv 2025

  44. [44]

    WorldPlay: Towards Long-Term Geometric Consistency for Real-Time Interactive World Modeling

    W. Sun, H. Zhang, H. Wang, J. Wu, Z. Wang, Z. Wang, Y . Wang, J. Zhang, T. Wang, and C. Guo, “Worldplay: Towards long-term geometric consistency for real-time interactive world modeling,” arXiv preprint arXiv:2512.14614, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  45. [45]

    Matrix-game: Interactive world foundation model.arXiv preprint arXiv:2506.18701, 2025

    Y . Zhang, C. Peng, B. Wang, P. Wang, Q. Zhu, F. Kang, B. Jiang, Z. Gao, E. Li, Y . Liuet al., “Matrix-game: Interactive world foundation model,”arXiv preprint arXiv:2506.18701, 2025

    work page arXiv 2025

  46. [46]

    Ca2-vdm: Efficient autore- gressive video diffusion model with causal generation and cache sharing,

    K. Gao, J. Shi, H. Zhang, C. Wang, J. Xiao, and L. Chen, “Ca2-vdm: Efficient autore- gressive video diffusion model with causal generation and cache sharing,” arXiv preprint arXiv:2411.16375, 2024

    work page arXiv 2024

  47. [47]

    Acdit: Interpolating autoregressive conditional modeling and diffusion transformer.arXiv preprint arXiv:2412.07720,

    J. Hu, S. Hu, Y . Song, Y . Huang, M. Wang, H. Zhou, Z. Liu, W.-Y . Ma, and M. Sun, “Acdit: Interpolating autoregressive conditional modeling and diffusion transformer,”arXiv preprint arXiv:2412.07720, 2024

    work page arXiv 2024

  48. [48]

    Test-Time Training Done Right

    T. Zhang, S. Bi, Y . Hong, K. Zhang, F. Luan, S. Yang, K. Sunkavalli, W. T. Freeman, and H. Tan, “Test-time training done right,”arXiv preprint arXiv:2505.23884, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  49. [49]

    Qwen-Image Technical Report

    C. Wu, J. Li, J. Zhou, J. Lin, K. Gao, K. Yan, S.-m. Yin, S. Bai, X. Xu, Y . Chen et al., “Qwen-image technical report,”arXiv preprint arXiv:2508.02324, 2025

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  50. [50]

    Opens2v-nexus: A detailed benchmark and million-scale dataset for subject- to-video generation.arXiv preprint arXiv:2505.20292, 2025

    S. Yuan, X. He, Y . Deng, Y . Ye, J. Huang, B. Lin, J. Luo, and L. Yuan, “Opens2v-nexus: A detailed benchmark and million-scale dataset for subject-to-video generation,”arXiv preprint arXiv:2505.20292, 2025

    work page arXiv 2025

  51. [51]

    Vbench: Comprehensive benchmark suite for video generative models,

    Z. Huang, Y . He, J. Yu, F. Zhang, C. Si, Y . Jiang, Y . Zhang, T. Wu, Q. Jin, N. Chanpaisitet al., “Vbench: Comprehensive benchmark suite for video generative models,” inIEEE Conference on Computer Vision and Pattern Recognition, 2024. 12

    work page 2024

  52. [52]

    Lora: Low-rank adaptation of large language models

    E. J. Hu, Y . Shen, P. Wallis, Z. Allen-Zhu, Y . Li, S. Wang, L. Wang, W. Chenet al., “Lora: Low-rank adaptation of large language models.” in International Conference on Learning Representations, 2022

    work page 2022

  53. [53]

    Unimatch: Universal matching from atom to task for few-shot drug discovery,

    R. Li, M. Li, W. Liu, Y . Zhou, X. Zhou, Y . Yao, Q. Zhang, and H. Chen, “Unimatch: Universal matching from atom to task for few-shot drug discovery,” arXiv preprint arXiv:2502.12453, 2025

    work page arXiv 2025

  54. [54]

    Q-Align: Teaching LMMs for Visual Scoring via Discrete Text-Defined Levels

    H. Wu, Z. Zhang, W. Zhang, C. Chen, L. Liao, C. Li, Y . Gao, A. Wang, E. Zhang, W. Sunet al., “Q-align: Teaching lmms for visual scoring via discrete text-defined levels,” arXiv preprint arXiv:2312.17090, 2023

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  55. [55]

    Fast-vqa: Efficient end-to-end video quality assessment with fragment sampling,

    H. Wu, C. Chen, J. Hou, L. Liao, A. Wang, W. Sun, Q. Yan, and W. Lin, “Fast-vqa: Efficient end-to-end video quality assessment with fragment sampling,” in European Conference on Computer Vision, 2022

    work page 2022

  56. [56]

    Arcface: Additive angular margin loss for deep face recognition,

    J. Deng, J. Guo, N. Xue, and S. Zafeiriou, “Arcface: Additive angular margin loss for deep face recognition,” inIEEE Conference on Computer Vision and Pattern Recognition, 2019

    work page 2019

  57. [57]

    Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution

    P. Wang, S. Bai, S. Tan, S. Wang, Z. Fan, J. Bai, K. Chen, X. Liu, J. Wang, W. Ge et al., “Qwen2-vl: Enhancing vision-language model’s perception of the world at any resolution,” arXiv preprint arXiv:2409.12191, 2024

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  58. [58]

    Musiq: Multi-scale image quality trans- former,

    J. Ke, Q. Wang, Y . Wang, P. Milanfar, and F. Yang, “Musiq: Multi-scale image quality trans- former,” inInternational Conference on Computer Vision, 2021. 13 ❓Upper ❓Lower ❓Full RawVideoDatabase ShotSegmentation OverallAssessmentHumanDetectionOptical-FlowEstimateStage1:GeneralCoarse-to-Fine Video FilteringStage2:Static-Dynamic VideoCaptioning Stage3:Fine-...

    work page 2021

  59. [59]

    Adaptive Reference Image Construction

    Fine-Grained Garment Image Extraction, and4. Adaptive Reference Image Construction. The overall curation pipeline is illustrated in Figure 8, and we detail each stage as follows:

    work page

  60. [60]

    These clips are then further divided into 3-5 second subclips, while discontinuous or overly short subclips are removed

    General Coarse-to-Fine Video Filtering.We collected a large set of raw videos from the Internet and filtered them in a coarse-to-fine manner usingShot Segmentation,Human Detection, Optical-Flow Estimation, andOverall Assessmentto retain only qualified videos: • Shot Segmentation.The raw videos are first processed with PySceneDetect to identify scene trans...

    work page

  61. [61]

    Static-Dynamic Video Captioning.For the filtered videos, we use the vision-language model (VLM) Gemini-3.1 to generate captions. Specifically, we adopt a static-dynamic decoupling strategy: • Static Caption.We prompt the VLM to focus on the static content in each video, includ- ing the scene layout, environmental atmosphere, human attributes (e.g., appear...

    work page

  62. [62]

    Since try-off is not always reliable in practice, we further introduce a VLM to verify the extracted garments

    Fine-Grained Garment Images Extraction.For each filtered video, we extract the initial frame and apply the image try-off model Qwen-Image-Edit [49] to extract corresponding garment images. Since try-off is not always reliable in practice, we further introduce a VLM to verify the extracted garments. In detail, for each extracted garment, the VLM performs a...

    work page

  63. [63]

    To improve training robustness, the garment worn by the person in the reference image should differ from the extracted garment

    Adaptive Reference Images Construction.In the final stage, we construct the reference image. To improve training robustness, the garment worn by the person in the reference image should differ from the extracted garment. We note that the garment information extracted in the previous stage may be incomplete, for example, including only the upper-body or lo...

    work page

  64. [64]

    She takes light steps forward, with her arms swinging naturally

    “A woman wearing a blue beret, earrings, and a watch stands on a floral garden path. She takes light steps forward, with her arms swinging naturally. Her gaze shifts from downward to focusing on the lens with a gentle smile, then smoothly transitions into a still pose, ensuring the movement is continuous and physically realistic.” For reproducibility, we ...

    work page

  65. [65]

    Initially facing the camera, she subtly shifts her body to the left and places her left hand into her pocket

    “A woman performs a series of poses in an indoor setting while holding a white handbag in her right hand. Initially facing the camera, she subtly shifts her body to the left and places her left hand into her pocket. She then moves her left hand to rest lightly on a black shelving unit behind her. Throughout the video, she maintains a friendly smile and st...

    work page

  66. [66]

    He maintains a steady forward gait, his arms swinging naturally to showcase the drape of the new garment

    “A man strolls along an outdoor brick path, wearing a brown turtleneck long-sleeved knit sweater paired with white shorts and beige sandals. He maintains a steady forward gait, his arms swinging naturally to showcase the drape of the new garment. The camera performs a smooth tracking shot, moving backward to keep him centered in the frame. Initially looki...

    work page

  67. [67]

    She initially tilts her head slightly to the side, then naturally shifts her gaze back to the lens with a soft smile

    “A young woman stands in a room, wearing a red short-sleeved t-shirt paired with a long floral skirt, with a red string bracelet on her left wrist. She initially tilts her head slightly to the side, then naturally shifts her gaze back to the lens with a soft smile. She performs a subtle turn to the left, causing the hem of the long skirt to sway with natu...

    work page

  68. [68]

    She walks forward with an elegant catwalk stride, her arms swinging naturally while her platform sneakers land steadily

    “A young woman walks near park flowers, wearing a blue zippered crop top and lace-up distressed denim shorts, accented with a white cap, necklace, and a bag featuring a teddy bear charm. She walks forward with an elegant catwalk stride, her arms swinging naturally while her platform sneakers land steadily. The camera performs a steady tracking shot, keepi...

    work page

  69. [69]

    She wears purple flower earrings and carries a pink woven bag on her shoulder

    “A young woman stands against a pink and blue background. She wears purple flower earrings and carries a pink woven bag on her shoulder. She walks forward with light steps, her arms swinging naturally, while the bag strap bounces slightly. She then tilts her head toward the camera with a bright smile and a natural blink. The movement is smooth and consist...

    work page

  70. [70]

    She wears a baseball cap with text and has one hand in her pocket

    “A young woman stands in a room filled with books and vintage items. She wears a baseball cap with text and has one hand in her pocket. She slowly lowers her hand from the cap, shifts her weight, and turns slightly to the right. Her gaze shifts from the lens toward the stack of books before turning back to blink and smile naturally. The movement is smooth...

    work page

  71. [71]

    She initially looks down in reflection, then gracefully turns her head to the left, gazing into the distance

    “In the video, a young woman slowly enters from the right side of the frame and stops near a table. She initially looks down in reflection, then gracefully turns her head to the left, gazing into the distance. The camera remains in a fixed position, capturing the scene through a transparent glass door, with subtle reflections on the glass shifting as she ...

    work page

  72. [72]

    She raises her left hand to touch and adjust her hair, tossing it over her shoulder while her arms move naturally in sync with her shifting posture

    “Captured from a static camera angle, a young woman with long, flowing black hair sways her body gracefully to a rhythmic beat. She raises her left hand to touch and adjust her hair, tossing it over her shoulder while her arms move naturally in sync with her shifting posture. Throughout the sequence, she maintains direct eye contact with the camera, exhib...

    work page

  73. [73]

    She slowly lowers her raised right arm and turns her body slightly to the left to showcase the skirt’s silhouette

    “On a lush tree-lined path, a woman wears a black and white checkered vest paired with a blue mini skirt featuring a cherry graphic, accented by a pearl necklace and white boots. She slowly lowers her raised right arm and turns her body slightly to the left to showcase the skirt’s silhouette. The camera orbits steadily around her in an arc. She shifts her...

    work page

  74. [74]

    She slowly lowers her raised right arm and takes a natural step forward to showcase the outfit

    “A woman in a blue cap and sunglasses stands by a white tiled wall, wearing a light grey multi-pocket hooded jacket, white wide-leg pants, and beige shoes, holding a brown bag with a bear charm. She slowly lowers her raised right arm and takes a natural step forward to showcase the outfit. The camera pans horizontally to the right; she turns her head from...

    work page

  75. [75]

    She walks steadily toward the camera, the long skirt’s hem swaying naturally and gracefully with her steps

    “A young woman stands outdoors wearing white headphones and sunglasses, dressed in a black short-sleeved T-shirt and a dark green button-front maxi skirt, carrying a black backpack with white socks and sneakers. She walks steadily toward the camera, the long skirt’s hem swaying naturally and gracefully with her steps. The camera pulls back smoothly to rev...

    work page

  76. [76]

    He walks forward with steady steps, his body swaying naturally to showcase the fit of the tank top

    “On a city street, a black-haired man wearing sunglasses is dressed in a black U-neck tank top paired with ripped blue jeans and a black belt, holding a brown leather bag in his right hand with a watch and bracelet on his wrists. He walks forward with steady steps, his body swaying naturally to showcase the fit of the tank top. The camera slowly zooms out...

    work page

  77. [77]

    She walks toward the camera with light catwalk steps, the layered hem swaying naturally

    “A young woman stands by the poolside with city buildings in the background, wearing a turquoise long-sleeved shirt and a white tiered ruffled long skirt, holding a small cream- colored handbag. She walks toward the camera with light catwalk steps, the layered hem swaying naturally. The camera slowly zooms out to reveal the full silhouette; her gaze shift...

    work page

  78. [78]

    She looks down initially, then raises her gaze to the camera while turning slightly to the left, allowing the skirt to drape naturally

    “A young woman stands in a minimalist gray indoor setting, wearing a white puff-sleeved blouse and a red phoenix-embroidered vest paired with a red patterned pleated skirt, with a thin bracelet on her left wrist. She looks down initially, then raises her gaze to the camera while turning slightly to the left, allowing the skirt to drape naturally. The came...

    work page

  79. [79]

    He slowly transitions from a leaning pose to a steady upright stance, balancing his weight on both feet to showcase the silhouette

    “In front of a white wall, a man wearing black-rimmed glasses holds a coffee cup, dressed in a tan sports bra and dark brown leggings with white sneakers. He slowly transitions from a leaning pose to a steady upright stance, balancing his weight on both feet to showcase the silhouette. The camera zooms out smoothly to capture the full outfit; he tilts his...

    work page

  80. [80]

    He moves his hands out of his pockets to his sides and walks forward toward the camera, the loose pant legs creating natural folds and swaying with each step

    “A young man wearing a baseball cap and black glasses stands before a dark rolling shutter, dressed in a white long-sleeved top and dark blue wide-leg trousers. He moves his hands out of his pockets to his sides and walks forward toward the camera, the loose pant legs creating natural folds and swaying with each step. The camera tracks him steadily; he ti...

    work page

Showing first 80 references.