Online fashion has a chronic problem: shoppers don’t trust the fit.

Now CATCHES thinks it can fix that—with physics, not guesswork.

At NVIDIA GTC, the company unveiled “RealFit,” a generative AI-powered virtual try-on system that promises something most fashion tech has struggled to deliver: accurate sizing, realistic fabric behavior, and a true-to-life preview of how clothes will actually look on your body.

Virtual Try-On, But With Physics

Unlike earlier virtual try-on tools that rely heavily on visual approximation, RealFit leans into simulation.

Built on NVIDIA’s CUDA and Omniverse platforms, the system combines generative AI with physics-based modeling to simulate how garments drape, stretch, and move. The result, CATCHES claims, is a “mirror-like” experience where shoppers can see how a piece fits—not just how it looks.

Here’s how it works:

• Users upload a photo and input body measurements

• The system generates a personalized digital twin

• Shoppers can try on garments virtually and toggle between sizes

• Fabric behavior is simulated based on real-world material properties

The first live deployment is already running on the AMIRI website, with more brand rollouts expected in the coming months.

Why Fit Is the Billion-Dollar Problem

Sizing uncertainty isn’t just a UX issue—it’s a revenue killer.

In some fashion categories, return rates exceed 50%, largely driven by poor fit. That creates a cascade of costs: reverse logistics, lost margins, and environmental impact.

RealFit is designed to tackle that head-on by giving shoppers confidence before they click “buy.” If it works as advertised, the upside is straightforward: higher conversion rates and fewer returns.

That’s a compelling pitch in a market where brands are under pressure to improve both profitability and sustainability.

The Tech Stack: GPUs, Digital Twins, and AI Models

Under the hood, RealFit is doing more than standard generative AI.

CATCHES spent two years building a GPU-accelerated simulation framework that models real fabrics—capturing weight, structure, and movement with high precision. The platform combines:

• Physics engines for fabric simulation

• Diffusion models for visual generation

• Vision-language and large language models for interaction

• NVIDIA’s accelerated computing stack for performance

The system runs on high-end infrastructure, including NVIDIA RTX and Blackwell GPUs, enabling millimeter-level fit accuracy and photoreal rendering.

In short: this isn’t just AI-generated imagery—it’s a hybrid of simulation and generation, which could mark a shift in how digital fashion experiences are built.

Industry Context: From Guesswork to “Physical AI”

Most virtual try-on solutions today—from startups to features embedded in platforms like Shopify—focus on visual overlays or size recommendations.

CATCHES is taking a different route: anchoring AI to physical laws.

That approach aligns with a broader trend in AI development, where companies are moving beyond probabilistic outputs toward systems grounded in real-world constraints—especially in areas like robotics, simulation, and design.

If successful, it could push the entire category forward from approximation to accuracy.

Early Backing and Luxury Interest

CATCHES has already raised $10 million from a mix of tech and luxury industry investors, including figures tied to LVMH and former executives from brands like Tommy Hilfiger.

That backing signals growing interest from high-end fashion, where fit, craftsmanship, and customer experience are central to brand value.

Luxury brands, in particular, may see RealFit as a way to replicate the in-store experience online—without sacrificing personalization.

What This Means for Retail and MarTech

RealFit sits at the intersection of retail, AI, and customer experience—squarely in MarTech territory.

For marketers and e-commerce teams, the implications go beyond sizing:

• Better conversion data: Understanding which sizes and styles resonate

• Personalized journeys: Tailoring recommendations based on body profiles

• Reduced churn: Fewer returns mean happier customers

• New engagement channels: Virtual try-on as a discovery experience

It also hints at a future where digital twins become a standard part of online shopping—especially as AI-driven personalization evolves.

Bottom Line

Virtual try-on has been around for years, but accuracy has always been the missing piece.

CATCHES’ RealFit is betting that combining generative AI with physics simulation can finally close that gap—turning a flashy feature into a functional tool.

If it delivers, it won’t just improve online shopping. It could fundamentally change how fashion is sold in the AI era.

Get in touch with our MarTech Experts.