CounterLook

The Problem

Virtual try-on has made it much easier to see what a product looks like on you, but it usually stops at visualization. It answers “What will I look like?” rather than “Is this actually the best choice for what I am trying to achieve?”

The Idea

We built CounterLook around a different idea: counterfactual visual simulation.

Instead of giving a user another static recommendation, CounterLook first simulates their chosen look using YouCam Apparel VTO. An agent then evaluates the result against the user's occasion and goals, identifies the primary visual trade-off, proposes alternatives, and tests those alternatives through virtual try-on.

How It Works

Understand → Simulate → Diagnose → Change → Re-simulate → Compare → Prove

In our primary demonstration, a user preparing for a job interview wants to look both professional and approachable. CounterLook simulates a structured black blazer, identifies silhouette rigidity as the main trade-off, tests three alternatives, and selects the strongest result.

The result is not simply “AI recommends this.” The system shows the original and alternative simulations and explains what changed and what was preserved.

Why YouCam

YouCam Apparel VTO is the visual simulation layer that makes the counterfactual loop possible. Each alternative can be tested on the user instead of being evaluated only through text or product metadata.

What We Learned

The most important design insight was that virtual try-on can be treated as an experimental environment, not merely a visualization endpoint. That opens the door to agentic systems that can test visual decisions rather than simply generate recommendations.

Consumer Value

CounterLook aims to reduce uncertainty in fashion decisions by helping shoppers understand the trade-offs between competing choices before they commit to one.