Phia
Virtual Try-On
Designed and shipped a scalable virtual try-on experience
Project Overview
Role
Lead Product Designer
Timeline
Aug 2026 (1 week)
Team
1 PM
1 iOS Engineer
Discipline
AI Design
Interaction Design
Retention Design
Virtual Try-On
With Virtual Try-on, users can generate a realistic preview of themselves wearing an item, helping them visualize how it looks and make more confident purchase decisions.
Making AI generation
The loading experience combines Phia's AI visual language with their chosen photo and item, making it feel personalized.
Context
47% of shoppers claim not being able to try-on something is a top reason online shopping falls short
User research revealed 47% of shoppers cited the inability to try on an item as one of the top reasons online shopping falls short of in-person shopping.
On Phia’s product pages that gap was evident with 88% of users dropping off in the PDP.
Problem
Product photos and descriptions weren't enough for users to make an informed decision
Ultimately, all users are asking themselves: “How will this look on me?”
"
Solution
Virtual try-on: A confidence-first feature built directly into the PDP, allowing users to experience an item
Virtual try-on would allow users to generate a realistic image of themselves wearing the item with minimal effort.
I designed the full journey around the goal of helping users build confidence to make informed, easy decisions.
Entry Point
Creating an entry point that draws just enough attention
Adding an entry point to an already-busy PDP meant balancing subtlety with discovery.
I used subtle motion with subtle colors to create enough visual salience to attract attention while maintaining hierarchy to the PDP.
I explored three styles, ultimately choosing the white variant for its balance of subtlety and brand expression.
Intro Screen
Building trust and curbing AI skepticism by showing, not telling
I demonstrated the feature’s versatility through real examples across product categories, helping users quickly understand its potential and trust the experience.
I chose products with distinct silhouettes across women’s and men’s clothing to demonstrate the feature’s versatility, from coats and dresses to pants and sweaters.
Image Preview
A streamlined path from image selection to try-on
For returning users, this preview replaces the intro screen entirely, taking them directly to their try-on and shortening time to value. A small preview of their chosen image keeps the streamlined experience feeling personalized and delightful.
Here, I also aligned with the engineering team to limit stored images to 5, balancing user choice with constraints around storage, performance, and implementation timelines.
Loading Screen
Designing AI latency to feel predictable, personalized, and delightful
I reused the AI visual language I established across other Phia features to create continuity with the broader product.
I also incorporated the user’s own photo and selected item to personalize the experience.
Image Results
Turning the generated image into a clear purchase decision
Using the virtual try-on feature signals high intent. In designing the results screen, I wanted to focus on reconnecting the experience to the purchase journey through the product and shop primary CTA.
The generated image becomes the primary focus, with secondary actions (regenerate, download, and change photo) kept accessible.
Impact
From concept to beta—designed to make virtual try-on easier to discover and return to.
I took the virtual try-on experience from early concept through iterative design and into beta. The final experience reduces friction for first-time users while giving returning users a faster path back to trying on products.
Reflection
Designing for familiarity at scale
I learned to design not just for a user's first interaction, but for the experience they might have after using a feature dozens of times.
From familiar button patterns and placement to the path users take to reach virtual try-on, every decision considered how the experience could become more intuitive with repeated use. The goal wasn't just to make the feature easy to learn, but easy to keep using.