<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Ikarus delta</title>
    <description>The latest articles on DEV Community by Ikarus delta (@ikarusdelta).</description>
    <link>https://dev.to/ikarusdelta</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4205177%2F0785ea3b-35eb-46a5-8383-c5faf1908fee.png</url>
      <title>DEV Community: Ikarus delta</title>
      <link>https://dev.to/ikarusdelta</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/ikarusdelta"/>
    <language>en</language>
    <item>
      <title>Building a Fast WebGL Product Viewer for Mobile</title>
      <dc:creator>Ikarus delta</dc:creator>
      <pubDate>Sat, 21 Nov 2026 15:59:27 +0000</pubDate>
      <link>https://dev.to/ikarusdelta/building-a-fast-webgl-product-viewer-for-mobile-46ap</link>
      <guid>https://dev.to/ikarusdelta/building-a-fast-webgl-product-viewer-for-mobile-46ap</guid>
      <description>&lt;p&gt;Interactive 3D on a product page sounds simple until you test it on a three-year-old Android phone over a weak connection. The model takes eight seconds to appear, the page stutters when you scroll and the battery drains. This post covers the practical decisions that make a product viewer feel instant on mobile, based on what we've learned shipping viewers and configurators for furniture, sauna and luggage brands.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Start with the right file format
For the web, glTF 2.0 (and its binary form, GLB) is the default. It's designed for runtime delivery: compact, fast to parse and supported by three.js, Babylon.js and &lt;code&gt;&amp;lt;model-viewer&amp;gt;&lt;/code&gt;.
Compress geometry with Draco or meshopt. Meshopt usually decodes faster on the client, while Draco can give smaller files. Test both on your real models.&lt;/li&gt;
&lt;li&gt;Budget your polygons per device
A model built for offline rendering can easily carry millions of triangles. On a product page, aim for a budget that keeps the frame rate steady on mid-range phones. Practical tactics:
Remove hidden geometry, such as the underside of a sofa frame or inside a closed cabinet.
Bake fine detail like stitching or wood grain into normal maps instead of geometry.
Ship LODs (levels of detail) and switch based on the device or zoom level.&lt;/li&gt;
&lt;li&gt;Treat textures as the real bottleneck
Textures often weigh more than the mesh. Use KTX2 with Basis Universal so textures stay compressed on the GPU, keep texture sizes proportional to how close the camera gets, and share texture atlases across parts where you can. For products with many materials, load only the material the shopper has selected, then fetch others on demand.&lt;/li&gt;
&lt;li&gt;Load after the page, not before it
The 3D viewer should never block the rest of the product page.
Show a static poster image in the viewer's slot immediately.
Lazy-load the viewer script and model when the slot scrolls into view or the user taps to interact.
Reserve the viewer's dimensions in CSS so the layout doesn't shift when it loads.
These steps protect your Core Web Vitals, which matter for both SEO and conversion.&lt;/li&gt;
&lt;li&gt;Keep the render loop honest
Don't render 60 frames per second when nothing is moving. Render on demand: redraw only when the camera moves, a material changes or an animation is playing. It saves battery and keeps scrolling smooth.&lt;/li&gt;
&lt;li&gt;Lighting that looks good and stays cheap
Image-based lighting with a small, pre-filtered environment map gives realistic reflections at low cost. Bake ambient occlusion into textures rather than computing it in real time. A soft contact shadow under the product adds a lot of realism for very little cost.&lt;/li&gt;
&lt;li&gt;Design the interaction for thumbs
One finger rotates; two fingers pinch to zoom; double-tap resets.
Limit the camera so shoppers can't fly inside the model or lose it off-screen.
Keep hotspots large enough to tap (44 px is a good minimum) and never cover the product with UI.&lt;/li&gt;
&lt;li&gt;Add AR without a separate app
On iOS, Quick Look uses USDZ; on Android, Scene Viewer uses glTF. Generate both from the same source model so the AR version always matches the viewer, including the selected material.&lt;/li&gt;
&lt;li&gt;Measure on real devices
Desktop DevTools won't tell you how a phone behaves. Test on a few real mid-range devices, track time-to-first-interaction and frame rate, and set budgets per product.
Wrapping up
A fast viewer is mostly about restraint: smaller models, smarter textures, lazy loading and rendering only when needed. Get those right and the 3D feels like part of the page rather than a heavy add-on.
If you'd rather not build this pipeline yourself, our team runs it as a managed service. See how the &lt;a href="https://www.ikarusdelta.com/solutions/3d-product-viewer" rel="noopener noreferrer"&gt;3D product viewer&lt;/a&gt; handles models, materials and AR for product pages.
Bonus: a simple performance checklist
GLB with Draco or meshopt compression
KTX2 textures, sized to camera distance
Hidden geometry removed; detail baked into normal maps
LODs for heavy products
Poster image shown first; viewer lazy-loaded
Render on demand, not every frame
Small pre-filtered environment map and a contact shadow
Camera limits and thumb-friendly controls
USDZ and glTF generated from the same source for AR
Tested on real mid-range phones
Handling many material variants
Configurators with hundreds of fabrics need extra care. Keep the base mesh shared and swap only material parameters and textures. Generate thumbnails for swatch pickers from the same materials, so what the shopper taps matches what they see. Cache recently used textures, and prefetch the next likely choices when the network is idle.&lt;/li&gt;
&lt;/ol&gt;

</description>
      <category>3dproductviewer</category>
      <category>3dconfigurator</category>
      <category>productconfiguration</category>
    </item>
    <item>
      <title>How a Web-Based 3D Product Configurator Works: Models, Rules and Pricing</title>
      <dc:creator>Ikarus delta</dc:creator>
      <pubDate>Sat, 24 Oct 2026 17:27:31 +0000</pubDate>
      <link>https://dev.to/ikarusdelta/how-a-web-based-3d-product-configurator-works-models-rules-and-pricing-5b32</link>
      <guid>https://dev.to/ikarusdelta/how-a-web-based-3d-product-configurator-works-models-rules-and-pricing-5b32</guid>
      <description>&lt;p&gt;If you've ever wondered what's behind a &lt;a href="https://www.ikarusdelta.com/guides/what-is-a-3d-product-configurator" rel="noopener noreferrer"&gt;3D product configurator&lt;/a&gt;, the kind that lets you change a sofa's fabric and see the price update, here's the architecture in plain terms.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;1. The 3D asset&lt;/strong&gt; : Everything starts with a model, usually converted from CAD into glTF/GLB for the web (and USDZ for iOS AR). The key decision is how you split it: every part a buyer can swap (legs, arms, bench, handle) becomes its own mesh or node, and every finish becomes a PBR material. Keep polygon counts and texture sizes tight. Mobile is where most buyers are.&lt;br&gt;
&lt;strong&gt;2. The option model&lt;/strong&gt; : Options are data, not code. A simple structure:&lt;br&gt;
product → option groups (size, fabric, legs) → option values → which nodes/materials each value shows or hides.&lt;br&gt;
Store it as JSON so merchandisers can change it without a deploy.&lt;br&gt;
&lt;strong&gt;3. Rules&lt;/strong&gt;: Rules stop impossible combinations: "heater X only with sizes ≥ 6 ft", "corner module requires two arm modules". Express them as constraints evaluated on every change, and disable (don't hide) invalid values so buyers understand why.&lt;br&gt;
&lt;strong&gt;4. Pricing&lt;/strong&gt;: Each value carries a price delta or a formula (per seat, per square foot). Recalculate on every change and show the breakdown. For B2B, the same engine produces a quote PDF or pushes line items to a CRM (visual CPQ).&lt;br&gt;
&lt;strong&gt;5. Rendering and UX&lt;/strong&gt; : Three.js, Babylon.js or model-viewer render the scene. Guided camera moves to the part being configured make complex products easier to understand. Lazy-load the 3D below the fold and show a poster image first to protect Core Web Vitals.&lt;br&gt;
&lt;strong&gt;6. AR&lt;/strong&gt;: Export the current configuration to GLB/USDZ and hand it to Scene Viewer (Android) or Quick Look (iOS). No app required.&lt;br&gt;
&lt;strong&gt;7. Integration&lt;/strong&gt; : On "add to cart", pass a configuration ID and the line items to the store (Shopify, WooCommerce, custom), and store the full configuration for production.&lt;br&gt;
That's the whole stack: asset, option data, rules, pricing, renderer, AR and integration. If you'd rather read a non-technical overview, here's a guide to what a 3D product configurator is and how buyers use it.&lt;/p&gt;

&lt;p&gt;(Written by the team at Ikarus Delta. We build and run configurators for manufacturers.)&lt;/p&gt;

</description>
      <category>webdev</category>
      <category>threejs</category>
      <category>ecommerce</category>
    </item>
  </channel>
</rss>
