The 3D Product Viewer Setup That Actually Sells Jewelry Online
A founder-voice guide to picking a 3D product viewer for online jewelry stores: material rendering, mobile load, AR scale, and the buyer questions that close.
If you sell jewelry online at $400+ average order and you have tested a 3D viewer that did not lift conversion, the problem is rarely whether to ship 3D. It is which kind. A viewer built for furniture or eyewear handles a ring badly. The category sells on the first reflection of light off a gold band; if the viewer cannot show that, no other feature matters. This guide covers what an online jewelry store actually needs from a 3D product viewer and the setup that closes the buyer's hesitation.
What a 3D product viewer for jewelry has to get right
Jewelry sells on three constraints that most generic 3D viewers miss. The first is material rendering: the way light bounces off brushed gold, the way a gemstone refracts at the edges, the way pavé reads at 12 inches versus 6. The second is scale: a buyer cannot judge whether a 6 mm ring band is right without seeing it on a hand. The third is mobile weight: most jewelry browsing happens on a phone, and a 12 MB GLB does not survive an LTE connection.
A viewer built for furniture or eyewear typically misses two of these because the constraints differ for those categories. A jewelry-specific setup ships material fidelity, AR or wrist-shot scale, and a sub-3 MB mobile bundle.
Why material rendering is the whole game
A buyer at $400 for a ring is not asking "what shape is this." They are asking "does this look like real gold." A flat product photo can answer the first question and not the second. The job of the 3D viewer is to answer the second.
The rendering pipeline that delivers this:
- PBR materials with measured albedo. Gold has a specific reflectance curve. A material tuned to a generic "yellow metal" preset reads as plastic. Use measured PBR values for 18k yellow gold, white gold, rose gold, and sterling silver.
- Studio-style HDR lighting. A jewelry viewer needs the same lighting brief as the product photographer. Two soft lights at 45 degrees with a fill from above, not a generic "outdoor" environment map.
- Anisotropic highlights for brushed finishes. A brushed gold band has a directional highlight that catches the eye when the buyer rotates the model. Standard PBR misses this; jewelry-specific shaders add it.
- Real refraction for gemstones. A diamond, a sapphire, an emerald all bend light differently at the table and the pavilion. The viewer should model the refraction index per stone, not fake it with a generic glass shader.
If you cannot tell a brushed finish from a polished finish in the viewer, the viewer is wrong for jewelry.
The buyer questions that decide the sale
A jewelry buyer at premium prices runs the same 4 questions before they buy:
- Does the gold tone match what I expected. Yellow gold versus rose gold versus white gold. The viewer has to render all three correctly.
- How does the stone catch light. A diamond cut grade affects how the stone "sparkles." The viewer needs refraction that reads honest.
- What does it look like on me. A 6 mm band on a 60 mm wrist looks different than on a 70 mm wrist. AR or a wrist-shot composite answers this.
- Can I see the back / the underside. Settings, prongs, and stamping matter on premium pieces. Photos rarely show this; rotation does.
A 3D product viewer that answers all four is rare. Most premium online jewelry stores ship a viewer that answers two of the four and lose the buyer at question three or four.
The 4Cs framework that buyers reference when comparing diamonds is documented at GIA's 4Cs education site. A viewer that supports a buyer doing 4Cs research feels honest; one that does not feels like a brochure.
What the AR or scale step has to do
AR for jewelry splits by sub-category. Rings and watches work well because the hand is steady and the camera focal length is forgiving. Earrings work less well because the lighting on the ear is hard to match. Necklaces and bracelets are the hardest because the AR has to track a moving body part.
The setup that works on Shopify and most headless storefronts:
- Rings. Native AR via iOS Quick Look (USDZ) and Android Scene Viewer (GLB). The buyer points the phone at their hand. The ring lands on the finger. Total file under 2 MB.
- Watches. Same as rings, with the wrist as the anchor. Adjust the strap length per buyer; some watch viewers let the buyer pick a wrist size.
- Earrings. AR is hit-or-miss on most phones. A photo composite ("see on a model") usually converts better than a shaky AR placement.
- Necklaces and bracelets. Skip AR. Use a 360 viewer plus a photo composite on a model. Save the AR build budget for rings and watches.
For brands that have invested in Gaussian splats, the in-page rotation comes from the splat and the AR placement comes from a separate small GLB. The splat vs 3D model decision guide covers when this combination beats one or the other on a category like jewelry.
Mobile bundle weight and what to compress
Most jewelry browsing happens on a phone. A buyer scrolling Instagram on LTE gives the viewer 2 to 3 seconds before they decide. A 12 MB GLB takes 9 seconds to load on LTE; the buyer is gone.
Aim for under 3 MB total bundle for the 3D viewer per product:
- Geometry compression. Draco mesh compression cuts the geometry size by 50 to 80%. Shopify-native viewers like Arty and LEVAR ship Draco support; verify your export uses it.
- Texture maps. Default exports often use 4K textures. For jewelry on mobile, 1K or 2K is usually enough. The buyer is looking at a ring at 400 pixels wide, not 4000.
- Lazy loading. The viewer should not load on first paint. It loads when the buyer scrolls near it. The hero photo loads first, fast, brand-correct.
The honest comparison of 3D viewer apps for premium Shopify stores walks through which app handles compression and lazy loading well.
What about a Gaussian splat for jewelry
For brands that can ship a Gaussian splat in addition to a model, jewelry is one of the categories where the splat wins. The reason is the same one rendering struggled with above: real material fidelity. A splat captures the actual reflection on the actual gold, with the actual lighting from the studio. A PBR material guesses; a splat measures.
The split that works in production:
- Splat. In-page rotation. Shows the real surface. Loads at 1 to 4 MB after compression.
- Small GLB. AR placement. Renders cleanly in iOS Quick Look and Android Scene Viewer. Sized to the product (rings ship lighter than watches).
- Hero photo. First paint. Fastest, brand-correct, sets the visual tone before either 3D asset loads.
The founder's guide to Gaussian splatting for ecommerce covers the splat capture pipeline. The photo count guide covers how many shots a jewelry splat needs (usually 6 to 8 plus macro detail).
Where to start if you have not shipped 3D yet
If your store has not tried a 3D viewer at all, start with one ring or one watch. Pick the product where buyers ask the most pre-purchase questions. Test the viewer on a real phone over LTE before you measure the conversion lift. Make sure the rendering matches the product photography. If the viewer makes the gold look like plastic, the viewer is wrong, not 3D.
If your store has tried a 3D viewer that did not lift conversion, the issue is usually material rendering or mobile weight, not the technology. Walk the 4-question buyer checklist above against your current viewer. The gap will be obvious.
The buyer is at $400+. The page is on a phone. The first reflection of light off a gold band is what closes the sale.