Mobile-VTON: High-Fidelity On-Device Virtual Try-On

Zhenchen Wan; Ce Chen; Runqi Lin; Jiaxin Huang; Tianxi Chen; Yanwu Xu; Tongliang Liu; Mingming Gong

cs.CVMar 1, 2026

Mobile-VTON: High-Fidelity On-Device Virtual Try-On

Authors: Zhenchen Wan, Ce Chen, Runqi Lin, Jiaxin Huang, Tianxi Chen, Yanwu Xu, Tongliang Liu, Mingming Gong

Organizations: University of Sydney · MBZUAI · University of Melbourne · Google

Abstract

Virtual try-on (VTON) has recently achieved impressive visual fidelity, but most existing systems require uploading personal photos to cloud-based GPUs, raising privacy concerns and limiting on-device deployment. To address this, we present Mobile-VTON, a high-quality, privacy-preserving framework that enables fully offline virtual try-on on commodity mobile devices using only a single user image and a garment image. Mobile-VTON introduces a modular TeacherNet-GarmentNet-TryonNet (TGT) architecture that integrates knowledge distillation, garment-conditioned generation, and garment alignment into a unified pipeline optimized for on-device efficiency. Within this framework, we propose a Feature-Guided Adversarial (FGA) Distillation strategy that combines teacher supervision with adversarial learning to better match real-world image distributions. GarmentNet is trained with a trajectory-consistency loss to preserve garment semantics across diffusion steps, while TryonNet uses latent concatenation and lightweight cross-modal conditioning to enable robust garment-to-person alignment without large-scale pretraining. By combining these components, Mobile-VTON achieves high-fidelity generation with low computational overhead. Experiments on VITON-HD and DressCode at 1024 x 768 show that it matches or outperforms strong server-based baselines while running entirely offline. These results demonstrate that high-quality VTON is not only feasible but also practical on-device, offering a secure solution for real-world applications. Code and project page are available at https://zhenchenwan.github.io/Mobile-VTON/.

Figures & tables

Figure 2: Training overview of Mobile-VTON . The left side illustrates our Feature-Guided Adversarial (FGA) Distillation process, where a high-capacity TeacherNet supervises two lightweight student networks: GarmentNet and TryonNet. GarmentNet learns to reconstruct the garment using feature-level distillation loss $\mathcal{L}_{\text{feature}}^{\text{G}}$ , while TryonNet is guided by both feature-level loss $\mathcal{L}_{\text{feature}}^{\text{T}}$ and adversarial loss $\mathcal{L}_{\text{GAN}}$ . The right side shows the main training pipeline of Mobile-VTON , where TryonNet and GarmentNet are jointly trained with garment-aware supervision. The Light-Adapter injects garment semantics, and the model is optimized using latent concatenation, temporal consistency loss $\mathcal{L}_{\text{cons}}$ , and reconstruction loss $\mathcal{L}_{\text{Diff}}$ to generate high-fidelity try-on images.

Figure 3: Qualitative comparison on VITON-HD and DressCode between our approach and server-side baselines. Zoom in for finer details.

Server-side Mask-based methods
Dataset	VITON-HD					DressCode Upper-body					Memory	On
Method	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ UN	KID↓ UN	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ UN	KID↓ UN	Usage(G)	Mobile
SD-VITON [ 41 ]	0.104	0.890	0.831	9.857	1.450	0.201	0.858	0.555	50.755	38.331	-	✗
LaDI-VTON [ 35 ]	0.166	0.873	0.819	9.386	1.590	0.157	0.905	0.773	22.689	2.580	7.63	✗
StableVITON [ 23 ]	0.142	0.875	0.838	9.371	1.990	0.113	0.910	0.844	19.712	2.149	13.84	✗
IDM–VTON [ 4 ]	0.102	0.868	0.875	9.156	1.242	0.065	0.920	0.870	11.852	1.181	18.47	✗

Table 1: Quantitative comparison of models trained and evaluated on VITON-HD and DressCode upper-body datasets. Bold and underline indicate the best and the second best results, respectively. “UN” denotes the unpaired setting. FID score is multiplied by 100.

Server-side Mask-based method
Dataset	VITON-HD In-the-wild
Method	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ UN	KID↓ UN
SD-VITON	0.174	0.857	0.804	17.151	5.326
StableVITON	0.201	0.863	0.848	12.983	2.804
IDM–VTON	0.137	0.832	0.931	11.373	1.133
Server-side Mask-free method

Table 2: Quantitative comparison of models on the VITON-HD In-the-Wild dataset. “UN” indicated the unpaired setting. FID score is multiplied by 100.

Figure 4: Comparison of virtual try-on results with and without the TCG and LC. Zoom in for more details.

TCG	LC	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ UN	KID↓ UN
✗	✗	0.119	0.874	0.798	11.231	3.136
✓	✗	0.111	0.879	0.805	10.814	2.744
✓	✓	0.088	0.893	0.833	10.211	2.023

Table 3: Quantitative comparison of models trained on the VITON-HD dataset with components. “UN” indicated the unpaired setting. FID score is multiplied by 100.

Figure 5: Qualitative ablation results on the DressCode upper-body test set. Removing key components degrades garment detail, structural consistency, or overall synthesis quality.

Method	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ (UN)	KID↓ (UN)
Ours	0.088	0.893	0.833	10.211	2.023
w/o TCG	0.108	0.881	0.804	10.914	2.939
w/o LA	0.168	0.827	0.704	13.652	5.091
w/o Distill	0.354	0.788	0.438	113.59	106.66

Table 4: Quantitative ablation results on the DressCode upper-body test set (UN setting). Each component contributes to improved perceptual quality and distribution alignment.

Figure 6: Qualitative comparison of try-on results fine-tuned on DressCode (a) and VITON-HD (b). DressCode-trained models produce sharper textures and more accurate alignment, especially in fine-grained regions (highlighted in red). Zoom in for details.

Dataset	LPIPS↓	SSIM↑	CLIP-I↑	FID↓ UN	KID↓ UN
VITON-HD
VITON-HD	0.102	0.877	0.818	10.558	2.685
DressCode	0.088	0.893	0.833	10.211	2.023
DressCode Upper-body
VITON-HD	0.058	0.932	0.832	13.527	2.473
DressCode	0.053	0.935	0.845	12.775	1.917

Table 5: Comparison of Mobile-VTON fine-tuned on VITON-HD and DressCode upper-body. “UN” indicates unpaired setting; FID is multiplied by 100.

Figure 7: Try-on results on the VITON-HD In-the-Wild test set, produced by Mobile-VTON . Zoom in for finer details.

Figure 8: Try-on results on the DressCode test set, produced by Mobile-VTON . Zoom in for finer details.

Figure 9: Try-on results on the VITON-HD test set, produced by Mobile-VTON . Zoom in for finer details.

Virtual Try-On

Virtual try-on generates images or videos showing people wearing selected garments while preserving appearance, fit, pose, and identity.

See all 55 papers

Jul 16, 2026

TAMF-VTON: Token-Wise Sparse Adaptation for Mask-Free Virtual Try-On

Diffusion-based virtual try-on (VTON) often relies on segmentation-derived masks that become unreliable when the reference garment differs substantially from the target person's clothing. Existing methods may also lose fine details and commonly focus on single-garment transfer. We present TAMF-VTON, an inference-time mask-free framework built on Qwen-Image-Edit-2509. Given a person image, one or more garment images, and a text instruction, it performs single- or multi-garment transfer without masks, parsing maps, pose representations, or spatial controls at inference. Its central component is token-wise sparse Mixture-of-Experts adaptation, which provides input-dependent capacity without predefined expert roles. Timestep-weighted frequency supervision regularizes high-frequency reconstruction, while adaptive offline data construction supplies cross-category and cross-silhouette training tuples; neither is used at inference. TAMF-VTON achieves the best scores on most reported paired and unpaired metrics among evaluated VTON baselines on the standard single-garment VITON-HD and DressCode benchmarks. Selected qualitative results illustrate multi-garment composition, subject-to-subject transfer, diverse scenes, body-shape variation, and illumination changes. With INT4 on an RTX 4090, TAMF-VTON generates 1K-resolution images under 15 seconds and 2K-resolution images under 30 seconds. Our project can be accessed via https://www.style3d.ai/ai-photoshoot/virtual-clothing-try-on.

Jie Wang, Qian He, Gaofeng He +2

State Key Lab of CAD & CG, Zhejiang University and Style3D Research, China · Style3D Research, China · State Key Lab of CAD & CG, Zhejiang University, China

Virtual Try-On Compositional Image Generation Diffusion-Based Image Editing

Ask PDF

Feb 16, 2026

OmniVTON++: Training-Free Universal Virtual Try-On with Principal Pose Guidance

Image-based Virtual Try-On (VTON) concerns the synthesis of realistic person imagery through garment re-rendering under human pose and body constraints. In practice, however, existing approaches are typically optimized for specific data conditions, making their deployment reliant on retraining and limiting their generalization as a unified solution. We present OmniVTON++, a training-free VTON framework designed for universal applicability. It addresses the intertwined challenges of garment alignment, human structural coherence, and boundary continuity by coordinating Structured Garment Morphing for correspondence-driven garment adaptation, Principal Pose Guidance for step-wise structural regulation during diffusion sampling, and Continuous Boundary Stitching for boundary-aware refinement, forming a cohesive pipeline without task-specific retraining. Experimental results demonstrate that OmniVTON++ achieves state-of-the-art performance across diverse generalization settings, including cross-dataset and cross-garment-type evaluations, while reliably operating across scenarios and diffusion backbones within a single formulation. In addition to single-garment, single-human cases, the framework supports multi-garment, multi-human, and anime character virtual try-on, expanding the scope of virtual try-on applications. The code is available at https://github.com/Jerome-Young/OmniVTON-PlusPlus.

Zhaotong Yang, Yong Du, Shengfeng He +5

School of Computer Science and Technology, Ocean University of China, Qingdao, China · School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing, China · Sanya Oceanographic Institution, Ocean University of China, Sanya, China +2

Virtual Try-On

Ask PDF

Mar 20, 2026

OmniDiT: Extending Diffusion Transformer to Omni-VTON Framework

Despite the rapid advancement of Virtual Try-On (VTON) and Try-Off (VTOFF) technologies, existing VTON methods face challenges with fine-grained detail preservation, generalization to complex scenes, complicated pipeline, and efficient inference. To tackle these problems, we propose OmniDiT, an omni Virtual Try-On framework based on the Diffusion Transformer, which combines try-on and try-off tasks into one unified model. Specifically, we first establish a self-evolving data curation pipeline to continuously produce data, and construct a large VTON dataset Omni-TryOn, which contains over 380k diverse and high-quality garment-model-tryon image pairs and detailed text prompts. Then, we employ the token concatenation and design an adaptive position encoding to effectively incorporate multiple reference conditions. To relieve the bottleneck of long sequence computation, we are the first to introduce Shifted Window Attention into the diffusion model, thus achieving a linear complexity. To remedy the performance degradation caused by local window attention, we utilize multiple timestep prediction and an alignment loss to improve generation fidelity. Experiments reveal that, under various complex scenes, our method achieves the best performance in both the model-free VTON and VTOFF tasks and a performance comparable to current SOTA methods in the model-based VTON task.

Weixuan Zeng, Pengcheng Wei, Huaiqing Wang +8

The Chinese University of Hong Kong, ShenZhen · Beihang University · KuaiShou

Virtual Try-On Diffusion Transformer Efficient Transformer Inference

Ask PDF

Abstract

Virtual Try-On

Explore similar work

TAMF-VTON: Token-Wise Sparse Adaptation for Mask-Free Virtual Try-On

OmniVTON++: Training-Free Universal Virtual Try-On with Principal Pose Guidance

OmniDiT: Extending Diffusion Transformer to Omni-VTON Framework