Home/ Projects/ Augmentlee

Augmentlee

A web-AR toolkit — a collection of APIs and plug-ins that drop into e-commerce sites and let shoppers view real 3-D models of products through their phone camera. No app install, no SDK.

RoleSolo build · architecture + dev
StackThree.js · Node.js · WebAR
StatusPrototype

The problem

Furniture, eyewear, sneakers — anything where shoppers want to "see how it'll look" — has the same online problem: a 2-D photo on a white background only gets you so far. Native AR apps solve this, but asking someone to install an app just to check whether a chair fits in their living room is a tall order.

Augmentlee asks: what if web is enough?

What it does

  • Drop-in plug-in — a small JavaScript snippet you embed in any product page.
  • Phone camera passthrough — taps into WebXR where it's available and gracefully falls back to a viewer-only mode when it isn't.
  • 3-D model API — handles upload, optimization, and CDN-friendly hosting of .glb assets.
  • Analytics hooks — events for "viewed in AR", "rotated", "added to cart from AR view" so the merchant can actually measure whether it's working.

How it's built

Three.js on the front-end for rendering — model loading, lighting, touch controls. Node.js on the server handles model ingest and the API surface. Everything is designed around being un-opinionated: the merchant brings the model, Augmentlee handles the rest.

The hardest part wasn't the AR. It was getting the lighting to look believable on a phone screen, outside, at noon.

Things I learned

Building for the web is full of polite lies — WebXR support is patchy, GLTF files come in roughly a thousand shapes, and most product photography is not lit anything like real life. Most of my time went into graceful degradation: making sure the product looked reasonable everywhere, even on devices that couldn't do real AR.

Where it goes next

I'd like to package this as a Shopify app and see whether real merchants would actually use it. The next milestone is a working install flow — drag-and-drop your model, paste a snippet, done.

Want to see the code?

The repository is on GitHub — issues and PRs welcome.

View on GitHub