80.lv published a breakdown of Ezequiel Grand’s career this week — an Argentinian 3D artist based in Los Angeles who has done commercial and concert-visual work for brands including Apple, Nike, Adidas and FIFA, and toured visuals for acts like Paul McCartney and Black Eyed Peas. The interesting part, for anyone who spends their day in a DCC rather than reading about one, is not the client list. It is the specific deadline that pushed him from modeling buildings by hand to building a tool that generates them.
The problem that actually forces the switch
Around 2018, Grand worked on visuals for Paul McCartney’s Freshen Up Tour, a job that needed several distinct environments — desert, underwater, more — built fast enough to hit a touring production’s schedule. Underwater meant FLIP simulations for the waterfalls and grain sims for sediment; desert meant terrain generation. And somewhere in the brief was a large number of buildings that all needed to look plausible, vary enough not to read as copy-pasted, and exist well before the schedule would tolerate modeling them one at a time.
That is the moment every environment artist eventually hits, on a smaller scale: the fortieth building on a block, the twentieth crate variant, the tenth near- identical rock formation. Model it by hand and the schedule eats you. Grand’s answer was to build a procedural house generator in Houdini instead — a network that exposes parameters for footprint dimensions, height, window count, roof design and other architectural features, so that variation becomes a matter of turning dials rather than opening a new file.
Why “expose the parameters” is the actual craft
It’s worth being precise about what separates a procedural generator from a pile of Houdini nodes that happen to make a house shape once. The difference is packaging: wrapping a network of SOPs into a proper digital asset (an HDA) with named, meaningful parameters on the outside, rather than leaving every control buried three subnetworks deep where only the person who built it can safely touch it.
Done that way, a generator earns its cost back three times over on a job like a touring visual package:
- Iteration speed. Changing “make it taller” or “give it a pitched roof” is a slider, not a remodel. On a schedule as tight as a concert tour’s, that difference is the whole reason the tool exists.
- Consistency across shots. Every building comes out of the same underlying logic, so a client note like “the roofs read too uniform” can be fixed once, in the generator, instead of hunted down across forty individual meshes.
- Reuse past the one job. A generator built to hit one deadline keeps paying rent afterward — the same network, or pieces of it, showing up in the next environment brief that needs plausible-but-varied buildings fast.
Anyone who already works procedurally will find that familiar. What the breakdown is actually useful for is the reminder of where the decision gets made: not at the concept stage, and not after the schedule has already slipped, but at the moment you notice you are about to model the same kind of thing more than a handful of times. That is the tell. Miss it and you spend the deadline modeling; catch it and you spend a fraction of the deadline building a tool, then generate your way through the rest.
Where this doesn’t apply
It’s not a universal argument for proceduralism over hand-modeling — a one-off hero asset that needs a specific silhouette is still better served by a modeler’s eye than a parameter set, and a generator built for a job it will only run once is wasted setup time. The judgment call is entirely about repetition: how many times, and how varied. Grand’s tour brief needed dozens of buildings that had to look different enough to sell a skyline, not one building that had to look exactly right. That’s the condition a generator is built for.
The breakdown also lists six of Grand’s own tips for building efficient procedural setups; rather than paraphrase them secondhand, I’d point anyone who works this way to the source below for the specifics straight from him. The shape of the lesson holds regardless of the exact list: procedural work pays for itself on repetition, not on principle, and the trick is noticing the repetition early enough to act on it.
Source: 80.lv, “Breakdown: How Houdini Is Reshaping Commercial Animation”