Revolutionizing online fashion with real-time, physics-based fit technology for accurate sizing, drape, and personalized shopping experiences.
CATCHES, an emerging force at the intersection of artificial intelligence and digital fashion, has unveiled a groundbreaking innovation that could redefine how consumers shop for clothing online. The company, which specializes in building both application and infrastructure layers for next-generation fashion experiences, announced the launch of its Generative AI-powered sizing solution known as “RealFit.” This new technology introduces a highly advanced, physics-based virtual try-on system designed to replicate real-world garment behavior with remarkable accuracy.
The launch took place during a major industry moment at NVIDIA GTC, signaling the importance of the collaboration between CATCHES and NVIDIA. Built on NVIDIA’s powerful CUDA platform and enhanced by Omniverse technologies, RealFit represents a major leap forward in bridging the gap between physical and digital retail experiences.
At its core, RealFit aims to solve one of the most persistent challenges in fashion e-commerce: uncertainty around sizing and fit. Online shoppers frequently struggle to determine how a garment will look and feel on their bodies, often leading to hesitation during purchase and, ultimately, high return rates. In some categories, returns due to poor fit can exceed 50%, creating significant inefficiencies and costs for retailers. CATCHES addresses this issue by offering a deeply personalized, photorealistic experience that allows users to visualize exactly how clothing will fit, drape, and move on their own bodies.
The process begins with the shopper providing their measurements along with a photograph. Using this data, RealFit generates a highly accurate digital twin of the individual. This digital representation is not just a static avatar—it is powered by a sophisticated simulation framework that understands real-world physics, including fabric weight, structure, and motion dynamics. Once the digital twin is created, users can try on garments virtually, switch between sizes, and observe how each variation behaves in real time. The result is an experience that closely mirrors trying on clothes in a physical fitting room.
The first live deployment of this technology is now available through AMIRI, a luxury fashion house known for its emphasis on craftsmanship, community, and innovation. The integration of RealFit into AMIRI’s digital platform marks a significant milestone, demonstrating how high-end fashion brands can leverage advanced AI to enhance customer engagement and personalization.
Mike Amiri, Founder and Creative Director of AMIRI, emphasized the alignment between the brand’s philosophy and the new technology. He highlighted how RealFit embodies the concepts of digital craftsmanship and individualized experience, enabling the brand to connect with customers in a more meaningful way—whether they are shopping in-store or online. This approach reinforces AMIRI’s commitment to innovation while preserving its identity rooted in artistry and community.
From a technological standpoint, RealFit is powered by a combination of cutting-edge tools and frameworks. It leverages NVIDIA’s CUDA and Omniverse libraries alongside the Newton Physics Engine to simulate realistic garment behavior. Additionally, the system incorporates custom-trained diffusion models and fine-tuned vision-language and large language models. These include AI systems based on NVIDIA’s Nemotron and Cosmos world foundation models, which help accelerate the creation of physically accurate generative experiences.
What sets CATCHES apart is its focus on grounding AI in the laws of physics. Rather than relying solely on pattern recognition or estimation, the system uses a GPU-accelerated simulation framework developed over two years. This framework was trained on real fabrics and materials, enabling it to produce a true 1:1 representation of garments, including their drape, texture, and movement. This level of precision ensures that the virtual experience is not only visually convincing but also functionally reliable.
The infrastructure supporting RealFit is equally impressive. CATCHES operates on a distributed, bare-metal architecture powered by high-performance GPUs, including NVIDIA RTX PRO 6000 Blackwell Server Edition, Hopper, and Blackwell Ultra systems. These are deployed at scale with edge inference capabilities provided by partners such as Nscale, Nebius, Boston Limited, and Scan Computers. This robust backend enables RealFit to deliver real-time, high-fidelity simulations to users around the world.
For brands, the benefits of this technology extend beyond improved customer satisfaction. By reducing uncertainty around fit, RealFit has the potential to significantly increase conversion rates and lower return volumes. This, in turn, reduces the costs associated with reverse logistics, one of the most expensive aspects of online retail. Moreover, the ability to offer a personalized and immersive shopping experience can strengthen brand loyalty and deepen the relationship between consumers and products.
CATCHES has also attracted significant investment, raising $10 million from a group of influential figures in the luxury and technology sectors. Among them are Antoine Arnault, Natalia Vodianova, Roy Chung, Dillon Erb, Gary Sheinbaum, and Sarah Willersdorf. Their backing underscores the growing belief that AI-driven, physics-based solutions represent the future of fashion commerce.
Antoine Arnault noted the importance of preserving the relationship between a garment and the individual wearing it, suggesting that technologies like RealFit play a crucial role in bringing that connection into the digital space. Meanwhile, Natalia Vodianova highlighted the broader implications of CATCHES’ approach, emphasizing how anchoring AI in physical reality marks a fundamental shift in how digital systems interact with the real world.
Looking ahead, CATCHES is already working with additional fashion brands, with several new deployments expected in the coming months. As adoption grows, RealFit could become a standard feature in online retail, transforming how consumers shop for clothing and how brands present their collections.
Ultimately, CATCHES’ RealFit represents more than just a technological innovation—it signals a paradigm shift in e-commerce. By combining AI with physics-based simulation, the company is creating a new category of digital experience that is both deeply personal and scientifically grounded. In doing so, it is not only addressing long-standing challenges in fashion retail but also paving the way for a more immersive, efficient, and connected future.
