AR Clothing Try on

Inspiration
Honestly, it started with a frustration most of us share — ordering three sizes of the same shirt just to send two back. Online fashion returns are a quiet disaster, both for shoppers and for the planet. Seeing a static product photo on a model who isn't me only does so much. I kept thinking: our phones already have great cameras and plenty of processing power, so why are we still guessing?

What it does
AR-Tryon is a webshop where you point your camera at yourself and see garments wrap onto your body in real time. It tracks your movements, fits clothing to specific body zones (head, upper, lower, shoes, full-body), and lets you swap colors, snap photos, and get a size recommendation based on your actual measurements. There's also a simple admin area for adding new products and viewing basic analytics.

How I built it
I focused on making the frontend experience fast and intuitive, with real-time rendering that feels natural in a phone's camera view. The backend is split intentionally: a lightweight public service handles analytics and product data, while a separate, isolated generation tool creates the 3D garments offline. This way, the heavy lifting doesn't slow down the shopping experience. The whole system is portable and can run anywhere.

Challenges I ran into
Getting clothing to align with a live pose was harder than it looked. The mirrored selfie view quietly broke my rotation logic, so shirts tilted the wrong way for a while. Pants kept sitting too high until I rethought the anchor points. Making a flat-looking garment follow shoulder and hip movement without appearing rubbery took a lot of fine-tuning. And keeping everything smooth on mobile meant making tough performance trade-offs.

Accomplishments that I'm proud of
The "drop a file and it just shows up" pipeline — name it correctly, and the product catalog updates automatically. The fact that you can stack multiple clothing items (hat + shirt + pants + shoes) and they all track together. The color-swapping feature on the t-shirt feels genuinely fun. And the decision to isolate the heavy generation work from the public storefront means scaling one doesn't drag down the other.

What I learned
AR fit is 20% math and 80% taste — the numbers can be "correct" and it still looks off. Mirroring is a trap; pick a coordinate convention early and write it down. Also: shipping a working demo beats shipping a perfect one, and isolating risky systems early saves a lot of pain later.

What's next
Real cloth simulation instead of stretched meshes, so fabric actually drapes naturally. Better depth perception — likely leaning into WebXR where available. Persistent measurements tied to a user account so size recommendations get smarter over time. Automating 3D garment generation from a standard product photo. And eventually, a proper checkout so the loop from "try" to "buy" closes inside the same session.

Built as a solo project to solve a real frustration. Would love to connect with others thinking about AR, retail tech, or reducing fashion returns.

Built With

• bun
• docker
• python
• typescript
• vercel