Think about how car buying used to work. You saw an ad in the newspaper or a commercial, you drove to a dealership, and a salesperson walked you through trims, features, and options in person, filling in whatever gaps existed in your head. You compared features, took test drives, and endlessly negotiated every detail.
That job has largely moved online. You find out a vehicle exists, you go to the manufacturer’s site, and you build it yourself: model, trim, color, wheels, interior, every option stacked on top of the last one. You select if you want to lease or buy it. Click a button and contact a dozen dealerships and by the time you actually walk into a dealership, you’re there to test drive and sign paperwork. The decision already happened somewhere else.
Which means the configurator isn’t a nice-to-have feature bolted onto a website. It’s the moment where somebody decides whether they’re actually going to spend a large amount of money on a vehicle. If it looks lame or is slow or is missing the option you wanted, you don’t call the dealership to ask about it. You just close that Google tab and leave. I’d bet cash money that a bad configurator quietly kills more sales than any single underperforming ad campaign, precisely because nobody logs the moment someone closes the tab.
So we’ve covered a bunch of topics so far in this series, but let’s get into that car configurator. Because you know that baby has 3D assets in there.
And to be clear, “automotive” in this series was never only cars. This applies to anything with an engine and a price tag steep enough that people research before they buy: motorcycles, ATVs, RVs, boats, camper vans. Everything that goes vroom, basically.
The numbers back this up wherever companies have bothered to measure it. Zodiac, the boat maker, built a configurator for its N-Zo line and watched sales of that model climb 50 percent over two years, with more than half of N-Zo sales now coming from customized builds instead of stock ones. Nautique replaced a pile of static catalog images with a configurator spanning twelve boat models. Yamaha runs configurators across its motorsports lineup, and Lee Somerhalder, the company’s Digital Marketing and Strategy Manager, has credited the tool with helping deliver “a significant lift in retail conversion.” None of that is a marketing department being generous. That is vroom vrooms driving off the lot.
So how does one of these things actually get built?
Most people have seen an online configurator, but very few have actually ever built one. When you switch a car from red to blue in a configurator, are you loading an entirely new 3D model? Or is something smarter happening under the hood?
It’s the smarter thing. You are not swapping models. You’re swapping materials and textures on one model that already contains every option.
Sketchfab, which is one of the platforms companies actually build these on, works this way through its Viewer API. You load a single base vehicle, then the configurator reassigns which texture set is active on which material slot in real time. Paint becomes a material swap. Wheels become a visibility toggle between a few pre-built options sitting invisibly inside the same file. New option gets added to the lineup next season? Someone adds a texture reference to a database. Nobody touches the model or rewrites the configurator.
This is the exact same concept behind the 150 percent USD file I walked through a few weeks ago (link above), just built for a browser instead of a full production pipeline. One file, with every option living inside it as a switchable variant, design and configurator, and marketing all pulling from the same source of truth instead of maintaining forty separate assets that inevitably drift out of sync.
After Sketchfab, the engine choice depends on what you’re optimizing for. Straight WebGL and three.js keep everything running client-side in the browser with no server cost per session. Unreal Engine and pixel-streamed Omniverse push rendering to a remote GPU and stream the video back, which buys you far higher visual fidelity at the cost of running (and paying for) server infrastructure for every person configuring a car at once. I covered this whole stack in more depth a few weeks back if you want the fuller picture. For what it’s worth, a real project also isn’t cheap or fast: industry estimates put a proper configurator build somewhere between five and fifty thousand dollars and four to eight weeks, and that’s before anyone scans a single material.
What you can fake, and what you actually cannot
Like everything in 3D…it’s not perfect. Real-time experiences for web configurators have limitations. Subsurface scattering, translucency, or any other visual element that requires secondary lighting calculations to look great don’t often work in these web viewers.
This comes up in the car’s glass, the interior leather, the seat foam under thin upholstery, tire rubber, even the taillight lenses on the vehicle you’re configuring all rely on some version of this effect to look convincingly real up close.
Real-time engines fake it with a handful of tricks, and every single one of them cuts a corner. Screen-space approximations only account for what’s currently visible on screen, which creates visible seams wherever geometry occludes itself. Pre-baked techniques assume the lighting and the surface curvature won’t change, which breaks the moment your configurator lets someone rotate the vehicle or switch the environment lighting from showroom to outdoor. None of these approaches are wrong exactly. They’re just closer to a clever illusion than an actual simulation, and a trained eye can tell the difference the moment they look closely.
Car paint has the exact same problem, and I already wrote a full deep dive on why this specific material is automotive’s hardest rendering problem. Real metallic and pearlescent paint is a stack: basecoat, millions of suspended flakes at random depths and orientations, a clearcoat on top that light has to pass through twice. Engines like Unreal approximate this with layered shading models that fake the interaction between coats well enough to look convincing in motion. What they’re not doing is simulating individual flakes scattering light at different depths the way an offline path tracer, or the CMF team validating that same paint against a physical plaque, actually would. That gap is exactly why the CMF renders that get boardroom sign-off, and the paint you see spinning in a browser configurator, are not held to the same bar, and probably shouldn’t be.
None of this means real-time configurators are bad at their job. It means their job is narrower than it looks. They’re built to sell a decision, not to survive the kind of close inspection a CMF team does before a $200 million interior program gets approved. Different tool, different bar, and confusing the two is where trust quietly erodes.
Where the walls are starting to move
The two technologies most likely to chip away at those limits over the next few years are moving in opposite directions.
Gaussian splats can already capture something close to photographic realism (if done well), including exactly the kind of light-scattering detail a hand-built real-time shader has to fake.
WTF is a 4D Gaussian Splat!?!?!?
A while back I wrote an article about a technology I was begrudgingly learning, and one of the running themes was that I’m getting worse at wanting to learn new things, not better. Every new piece of software, every new buzzword, and I catch myself getting frustrated before I’ve even given it a fair shot.
The catch is that a splat today is fundamentally a frozen capture of one specific object in one specific configuration. You can’t yet select “blue” and have a splat regenerate itself in a different paint color the way a configurator needs to. Standards work is underway to change that: Khronos ratified a Gaussian splat extension for glTF in early 2026, and browsers are shifting from WebGL workarounds to native WebGPU rendering, which is the kind of infrastructure work that tends to precede an actual product wave rather than follow it. Nobody’s cracked configurable splats yet. Photoreal and customizable are still two different technologies pointed at each other.
Meanwhile, the render-engine side of this gap has been closing from the other direction. I wrote a while back about how real-time and path-traced rendering used to be two completely different categories of image quality, and how NVIDIA’s RTX hardware and smarter denoising have been steadily narrowing that gap. Real-time ray tracing in a browser configurator is still a heavier lift than most companies want to pay for today, but “still” is doing a lot of work in that sentence. Every year the honest answer to “can we do this in real time” covers a little more ground than it did the year before.
I mean, just look at these little vroom vrooms!
The takeaway
A configurator is the single moment where a piece of 3D software has to be convincing enough to move real money, in real time, on hardware you don’t control. That’s a much harder bar than a marketing render sitting still under lighting someone spent a week perfecting. The parts that still fake it, the flake in the paint, the give in the leather, aren’t a failure of the artists building these experiences. They’re the actual edge of what real-time rendering can do right now, and watching that edge move outward year over year is one of the more interesting things happening in this entire industry.
Next week we start to wrap this series by looking at the people and events actually worth following in automotive 3D, the ones putting this stuff into practice before we move on to what’s next.
Who Am I?
New 3D Artist Merch!
New 3D Artist swag is here!
We ran a design contest inside the community, and Randy Mills took it home with a CMYK sweatshirt design that’s honestly too good not to wear.
Grab yours today.
3D News
Software & Tools
Brand-New Lumen Mode Is Coming To Unreal Engine 6 - 80 Level
Foundry Releases Nuke 17.1 - CG Channel
V-Ray Expands Gaussian Splatting Support to Cinema 4D - Radiance Fields
Maxon Releases Autograph 2026.1 - Maxon
Check Out Promising New Blender Retopology Tool Smart Remesh - CG Channel
Apple Unveils New M5 Ultra and M6 Processors - CG Channel
Get Free VFX Image Viewer and Video Playback Tool DJV 3.5 - CG Channel
AI + 3D
Meshy 7 Puts a Number on Image-to-3D Alignment - Unite.AI
Postshot 1.1.69 Adds SPZ4 Support and HTML Export for Gaussian Splats - CG Channel
Slapshot Launches Comfy Nodes for Its AI VFX Toolkit - Televisual
Topaz Labs’ Hyperion 2.5 Brings AI SDR-to-HDR Conversion to Astra - CG Channel
Industry News
Registration Is Now Open for SIGGRAPH Asia 2026 - ACM SIGGRAPH Blog
Watch Framestore’s VFX Breakdown for War Machine - Befores & Afters
See DNEG’s Breakdown of Its Period-Accurate VFX for Star City - CG Channel
See Troll VFX’s Breakdown for Hurricane/Shark Survival Drama Thrash - CG Channel
3D Job Spreadsheet
Link to Google Doc With A TON of Jobs in Animation (not operated by me)
Hello! Michael Tanzillo here. I am the Senior Director of Business Development at KitBash3D (Greyscalegorilla, Artstation, Sketchfab). Previously, I was the Head of 3D Tech Artists at Adobe and a Senior Artist on animated films at Blue Sky Studios/Disney with credits including three Ice Age movies, two Rios, Peanuts, Ferdinand, Spies in Disguise, and Epic.
In addition to my work as an artist, I am the Co-Author of Lighting for Animation: The Visual Art of Storytelling and the Co-Founder of The Academy of Animated Art, an online school that has helped hundreds of artists around the world begin careers in Animation, Visual Effects, and Digital Imaging. I also created The 3D Artist Community and this newsletter.
www.michaeltanzillo.com
Free 3D Tutorials on the Michael Tanzillo YouTube Channel
Thanks for reading The 3D Artist! Subscribe for free to receive new posts and support my work. All views and opinions are my own!