Tstars-Tryon 1.0: Robust and Realistic Virtual Try-On for Diverse Fashion Items
Mengting Chen, Zhengrui Chen, Yongchao Du, Zuan Gao, Taihang Hu, Jinsong Lan, Chao Lin, Yefeng Shen, Xingjian Wang, Zhao Wang, Zhengtao Wu, Xiaoli Xu, Zhengze Xu, Hao Yan, Mingzhou Zhang, Jun Zheng, Qinye Zhou, Xiaoyong Zhu, Bo Zheng
2026-04-22
Summary
This paper introduces Tstars-Tryon 1.0, a new system for virtually trying on clothes. It's designed to be a practical, high-quality solution for real-world use, like on an app where many people are using it at once.
What's the problem?
Existing virtual try-on technologies weren't quite ready for everyday use. They often struggled with realistic results in challenging situations like unusual poses, bad lighting, blurry images, or when people wanted to combine multiple images. Plus, they were often too slow to provide a smooth experience for users.
What's the solution?
The researchers created Tstars-Tryon 1.0 by building a complete system, not just a single model. This included a new model design, a way to manage lots of image data, a strong technical setup, and a specific training process. This system focuses on creating very realistic images, handling difficult conditions well, allowing users to combine multiple reference images, and working quickly enough for real-time use. It can handle different types of clothing and even change the person's appearance and background.
Why it matters?
This work is important because it makes virtual try-on technology actually useful for a large number of people. It’s already being used by millions of users on the Taobao App, showing it can handle a real-world workload. The researchers also released data to help others improve this technology further, which could lead to even better virtual shopping experiences.
Abstract
Recent advances in image generation and editing have opened new opportunities for virtual try-on. However, existing methods still struggle to meet complex real-world demands. We present Tstars-Tryon 1.0, a commercial-scale virtual try-on system that is robust, realistic, versatile, and highly efficient. First, our system maintains a high success rate across challenging cases like extreme poses, severe illumination variations, motion blur, and other in-the-wild conditions. Second, it delivers highly photorealistic results with fine-grained details, faithfully preserving garment texture, material properties, and structural characteristics, while largely avoiding common AI-generated artifacts. Third, beyond apparel try-on, our model supports flexible multi-image composition (up to 6 reference images) across 8 fashion categories, with coordinated control over person identity and background. Fourth, to overcome the latency bottlenecks of commercial deployment, our system is heavily optimized for inference speed, delivering near real-time generation for a seamless user experience. These capabilities are enabled by an integrated system design spanning end-to-end model architecture, a scalable data engine, robust infrastructure, and a multi-stage training paradigm. Extensive evaluation and large-scale product deployment demonstrate that Tstars-Tryon1.0 achieves leading overall performance. To support future research, we also release a comprehensive benchmark. The model has been deployed at an industrial scale on the Taobao App, serving millions of users with tens of millions of requests.