A Jewellery Ring Configurator, Taken Apart: What It Does and What Building One Involves

A customer likes one of your rings. They want to see it with an oval stone instead of a round one, a wider band, and white gold rather than yellow. You can photograph three variations, or you can describe them. Neither answers the question they are actually asking, which is what their combination looks like.

We built ARTLAND Atelier to make that question answerable in a browser. A visitor shapes a ring, turns it in three dimensions, opens the setting up to see how it is built, adds an engraving, and shares the result as a link. Every choice feeds one connected model, so the design stays there to be examined from any angle rather than resetting to a photograph.

The part that matters technically: the ring geometry is generated from its parameters. Changing a dimension rebuilds the affected parts. There is no library of pre-made models with one file per slider position, which is the difference between adding a variant and rebuilding for it.

This is a studio demonstration, not a product for sale. We are showing it because the interesting question for a jewellery brand is never “can it spin?” — it is what happens between a visitor choosing an option and your sales team receiving something they can act on. What follows is the walkthrough, feature by feature, with the decisions behind each one and the limits stated where they exist.

Watch it move first

A guided run through the application, from the first stone choice to a finished design. The figures below let you stop on individual features at your own pace.

Request access to the live demo→

Start with the stone, and watch the setting follow

The Stone tab holds shape, material, carat and cut-quality controls together, because in practice nobody changes one of them in isolation. Nine shapes are built in: round, oval, cushion, pear, marquise, princess, radiant, emerald and Asscher. Fifteen stone materials run from diamond and moissanite through sapphire, ruby, emerald, aquamarine, tanzanite and further.

Changing the shape does more than swap the face of the ring. A different outline produces different stone geometry, and the setting is rebuilt around its actual dimensions. A pear presents a different silhouette and a different support problem than an oval — which is far easier to understand by rotating the ring than by looking at two front-on photographs.

The carat slider runs from 0.25 to 5.00 ct, and the stone’s dimensions and the head around it update as it moves. That gives a visitor a way to judge the scale of a stone against the band, rather than trying to picture what 1.8 ct means.

Cut quality is a visible change, not a badge

The cut-quality presets change the stone’s proportions, including the angles that decide how light travels through it. The renderer traces refraction, internal reflection and dispersion — the coloured flashes that move as the stone or the lighting moves — through the stone’s actual facet planes. Switching between Ideal, Shallow and Deep is therefore a real visual comparison rather than a label change, and brilliance visibly collapses when the pavilion angle leaves the ideal range.

The light-return meter beside it is an indicative model value. It exists to explain a design change. It is not a grade for an individual gemstone, and no configurator should imply otherwise. A retail build would connect the visual configuration to the brand’s real stones and their verified certification data.

A setting and a band you can take apart

The Setting controls cover prong and bezel styles, three, four, six or eight prongs, adjustable head height, and an optional halo. These are structural changes with visible consequences: raising the head changes the side profile, and a bezel wraps the stone where prongs leave it exposed.

The exploded view separates stone, setting and band. It is the single most useful control for a customer who cannot yet picture how a ring is assembled — and for a designer who needs to check that a piece which looks convincing head-on also resolves from the side.

The band decides the character more often than the stone does

The Band tab carries eight profiles — court, D-shape, flat, flat-court, knife edge, bevelled, concave and half-round — with straight, tapered, cathedral and Euro shanks. Width adjusts from 1.5 to 8 mm. Ring size is shown in US sizing with the corresponding UK and EU values and the inner diameter, because the same customer will be quoted in all three.

A 4 mm court band and a 2 mm knife edge give the same stone completely different weight and character. Keeping both in the same scene, under the same light, makes that comparison honest in a way two product photographs shot months apart never are. For a brand, it is also a practical starting point for deciding which widths and combinations you actually want to offer.

Ten metals under the same light

Ten metal options: 14k and 18k yellow, rose and white gold, platinum, palladium, sterling silver and titanium, each with measured spectral response rather than a tinted approximation — including the rhodium behaviour that makes white gold read the way it does. Four finishes — high polish, satin brush, matte and sandblast — change how the surface answers the light.

That distinction matters more than colour. A polished gold band throws hard, bright reflections; a matte finish spreads them. Because the configurator changes the material’s behaviour and not just its hue, a visitor can judge a finish on the shape they have actually chosen.

The metal weight shown is calculated from the generated geometry and the configured alloy density, so changing width or alloy changes the readout. Be clear with customers about what that is: a model-derived estimate for discussion. Supplier alloys, manufacturing allowances, finishing loss and your pricing rules belong in the commercial specification, not in a viewer.

An engraving you can turn the ring to read

A personalisation field is worth very little if the customer cannot see the result. In the Details tab the text appears inside the band, in script, serif or sans, and the close view shows the lettering following the curve of the inner surface.

In this build the engraving changes the band geometry itself and travels with the exported mesh, rather than being painted on as a texture. The demonstration keeps the visible controls to the text and the three fonts; size, depth and position exist behind an expert view rather than in the main panel. That split is the point. For a real collection, the limits customers see should mirror the lettering, character counts and engraving process your workshop can genuinely deliver — an interface that allows forty characters where the bench can cut twenty is a complaint waiting to happen.

Accents work the same way. The panel a visitor sees offers pavé coverage — none, half or full — with the bead geometry generated along the band and the accent count and total accent carat reported. Channel setting is built too, but it sits in the same expert view. A commercial version decides that split deliberately: expose the choices your customers actually make, and keep the specialist adjustments out of the buying journey.

Move to inspect, stop to see it resolve

Visitors can orbit the camera, move into a macro view, or let the turntable run. Lighting and backdrop are controls in their own right: a bright environment reads the silhouette, a dark one reads the reflections, and dragging the key light moves the highlights across metal and stone.

While anything is moving, the renderer prioritises responsiveness. When the visitor stops, progressive refinement accumulates samples into a finished image — soft shadows, edge detail and depth of field resolving over a second or two. The pause becomes the moment the last change is actually judged. When the image has converged the render loop stops entirely rather than redrawing an identical frame, which is what keeps a page like this from cooking a laptop battery.

Scene controls include floor reflection, ambient occlusion, depth of field and caustics where the device supports them, with quality tiers chosen from the hardware on first run. One honest note: the most detailed desktop presentation is not a promise of identical performance on every phone. Any production deployment needs testing on the devices your audience actually carries, and that testing belongs in the project plan rather than in a demo.

From “something like this” to a specific design

The Share action encodes the configuration into a link. Opening it restores the design, so a customer can send the ring to a partner, or bring the exact combination into a conversation with your studio. That single feature is the difference between an enquiry that says “something like the third one on your website” and one that arrives with the design attached.

The application also produces a design specification view and several exports, each for a different job:

• PNG renders record the appearance of a configuration, with a 4K-width option and a transparent background where you need to composite it.
• The specification sheet summarises the selected design and prints or saves through the browser.
• Project JSON stores the editable configuration so it can be reopened later.
• STL and OBJ export the geometry, subject to the application’s manifold-validity checks.
• Turntable and light-orbit video capture a moving presentation in supported browsers.

Two limits worth stating plainly, because the temptation to blur them is strong. The specification is a design record, not a gemstone certificate. And a watertight-mesh check is useful preparation for a geometry handoff — it is not approval for manufacture. Stone fit, tolerances, minimum sections and the chosen casting process still need a jeweller’s review, and any brand that treats a validated STL as a green light will find that out expensively.

The enquiry step can carry the specification into an email to the studio. Connecting that to a retailer’s CRM, live stock, pricing or checkout is a separate integration, and it should be scoped around the systems you already run rather than bolted on afterwards.

The product rules matter as much as the picture

Everything above is the visible half. The half that decides whether a configurator earns its cost is the rule set underneath it: which stones fit which settings, which widths exist in which alloy, what is stocked, what is made to order, and what needs your approval before anyone can buy it. Those answers determine what a customer is allowed to select and what your sales team receives at the other end.

None of that is a rendering problem, and none of it is solved by the part you can see. We wrote about the money side separately in our 3D product configurator cost breakdown — the short version is that the per-SKU asset work is what buyers underestimate, not the platform.

Start with one journey that ends somewhere

A focused configurator for one collection, answering one clear question — “what will my combination look like?” — is a better first build than exposing every adjustment the software can make. Define that journey and the step that follows it, whether that is an enquiry, a quotation or a purchase. Everything else can come later, and most of it will turn out not to be needed.

What we would pin down before building yours

• The collection and its rules. Product families, permitted combinations, size ranges, materials, and the reference material needed to represent them accurately.
• The job the customer is trying to finish. Comparing finishes, personalising a piece, requesting a quotation, or buying a configured product outright — these lead to different builds.
• What has to travel with the design. A shared link, a saved project, a specification, customer details, or a connection into your sales systems.
• The delivery environment. Your site, the devices your customers use, hosting, load-time expectations, and who maintains the product options after launch.

Those decisions are what separate a material selector from a configurator that generates geometry and feeds a sales process. They also make the visual work purposeful rather than decorative.

One more thing worth planning early: the same models can serve more than the configurator. Here, the same parametric output produced the interactive experience, the product stills in this article and the demonstration film. Deciding those outputs together is how a product looks consistent across its page, its animation and its sales material — which is the same argument we make for jewellery CGI generally, and for interactive presentation work more broadly.

Questions we get before a configurator brief

See the working configurator

The demo runs on request rather than in the open. Ask below and we will send you a login, usually within a business day.

Build one around your collection

Everything above was built for a demonstration ring. The same approach — generated geometry, real computed weights, rules that match what your workshop can actually make — applies to a collection with your stones, your settings and your permitted combinations. What changes is the rule set, and that is a conversation about your products rather than about software.