AR and 3D models for Shopify
Shopify supports 3D product models natively. You need a GLB file for the web viewer and Android AR, and a USDZ for iOS AR Quick Look. The technical work is not the format — it is getting a model small enough to load quickly on a phone while still looking like the product people are about to buy.
The two formats and what each is for
- GLB — the web standard. Powers the on-page 3D viewer and Android AR. One binary file containing geometry, materials and textures
- USDZ — Apple’s format. Powers AR Quick Look on iPhone and iPad, launched straight from Safari
- You need both. They are exports of the same model, not different models
Weight is the whole problem
A model that looks perfect and takes eight seconds to appear has failed, because the customer has already scrolled. The offline render of a product might be ten million polygons with 4K textures per material; the web version of the same product has to be a small fraction of that and still hold up at the size it is actually viewed. That reduction is skilled work — decimating geometry while protecting silhouettes, baking detail into normal maps, atlasing textures, and compressing with Draco or Meshopt. It is not an export setting.
What good looks like
- Well under a few megabytes total, textures included
- Silhouette preserved — that is what the eye reads at thumbnail size
- Detail baked into normal maps rather than carried as geometry
- Correct real-world scale, or AR placement will be visibly wrong
- Sensible origin and orientation so the model sits on the floor, not through it
- PBR materials that survive being lit by an unknown environment
Where AR actually helps
Not everywhere. It earns its keep for products where size and fit in a real space are the main hesitation — furniture, appliances, large decor, anything a customer is unsure will fit. For small items it is a novelty that adds weight to the page. The honest test is whether a customer has ever emailed asking how big something is; if they have, AR answers it, and if they have not, spend the money on better imagery instead.
Getting them from a render project
If your products are being modelled for packshots anyway, the marginal cost of web-ready exports is small compared with building them later from scratch — but only if it is planned. Say at quoting stage that GLB and USDZ are needed, because a model built knowing it must run in a browser is topologised and textured differently from one that only has to render.
Questions
What file formats does Shopify need for 3D products?
GLB for the on-page 3D viewer and Android AR, and USDZ for iOS AR Quick Look. Both are exports of the same source model.
How big should a 3D product model be?
As small as it can be while still reading correctly at the size customers view it — well under a few megabytes including textures. Weight matters more than polygon count, and textures are usually the larger share.
Can you make AR models from our existing renders?
Yes, though the render model normally has to be rebuilt or heavily optimised. Detail is baked into normal maps, geometry is reduced while protecting the silhouette, and textures are atlased and compressed.
Does 3D on a product page actually help conversion?
It helps most where size or fit is the main hesitation — furniture, appliances, large items. For small, well-understood products the gain is small and the page weight is real.
Do we need a separate model for each colourway?
Usually not. One model with swappable materials covers a range, which keeps total page weight down and means a new colourway does not mean a new upload pipeline.