Stop Asking About Portfolios: The Interview Question That Reveals a Real Game Artist

Your portfolio review is a lie. Stop relying on it.

Every game art interview starts the same way: "Walk me through your portfolio." You get a rehearsed tour of pretty renders, and you learn almost nothing about whether this artist can ship. The portfolio shows the final image, not the thousands of decisions that got there. It's a curated highlight reel, and in a real production, the highlight reel doesn't matter. What matters is whether the artist can hit a triangle budget, build a clean UV layout, and hand off an asset that doesn't melt the engine.

I've interviewed dozens of artists. The ones who stumble aren't the ones with weak portfolios—they're the ones who can't answer the question: "Walk me through your LOD chain for that character." That question, not the portfolio, separates the artists who understand the pipeline from the ones who just know how to sculpt.

What does "LOD strategy" even mean?

Level of detail (LOD) is the practice of swapping a mesh for a simpler version as it gets farther from the camera. It's one of the oldest tricks in real-time rendering, proposed by James H. Clark in 1976. The idea is simple: don't render 100,000 triangles for a tree that's 200 meters away.

In a typical game asset, you'll have an LOD0, the high-detail mesh you see up close, and then progressively cheaper versions. Unity's guidance is to reduce triangle counts by roughly 50% between successive LODs. So if your LOD0 is 10,000 triangles, LOD1 should be around 5,000, LOD2 around 2,500, and so on. A technical artist might spec an LOD0 character at roughly 15,000 triangles with 2048x2048 textures and 2-3 draw calls, scaling down to LOD3 at 800 triangles. That's the kind of concrete spec an artist should be able to discuss.

When you ask about LODs, you're not just asking about a technical checkbox. You're asking about the artist's understanding of the entire pipeline: how they decide what detail matters, how they retopologize, how they bake, and how they think about performance.

The misconception: "Nanite killed LODs."

Here's the myth I hear constantly: "Unreal Engine 5's Nanite removes the need for LODs, so why bother asking about them?" Yes, Nanite can stream and scale film-quality meshes with hundreds of millions of polygons in real time, and it removes the need to hand-author LODs or bake normal maps. That's true.

But Nanite is a Unreal Engine feature. It's not universal. Unity doesn't have Nanite. Godot doesn't. And even in Unreal, Nanite has limitations—it's not ideal for skinned meshes or for mobile. The industry still ships games on mobile, and mobile is where LODs matter most. The fact that one engine has a solution for one platform doesn't mean LODs are dead.

An artist who says "Nanite killed LODs" is telling you they don't think beyond the demo. An artist who can explain when to use Nanite and when to hand-author LODs is telling you they understand the platform constraints. That's the filter you want.

The pipeline question that reveals everything

Here's the question I now ask every candidate: "Take me through your high-poly-to-low-poly workflow for a character. Where do you make the decisions about edge flow and topology?"

The answer tells you more than any portfolio. A real artist will describe blocking out the shape, sculpting a dense high-poly mesh in ZBrush, retopologizing a low-poly mesh, unwrapping UVs, baking the detail, and texturing. They'll mention edge flow around joints, eyes, and mouth—the places where animation deformation happens. They'll talk about why they prefer quads for deformation, as the standard pipeline does, and how they use tools like ZBrush's ZRemesher for quick retopology on props but manual tools like Maya's Quad Draw for characters that need to deform.

An artist who can't articulate this is either junior or has only worked in a bubble where someone else did the technical work. You don't want that person on a team that ships.

Compare the candidates: portfolio vs. pipeline talk

The portfolio is still useful—it shows you whether the artist has taste. But taste without technical skill is just a hobbyist. The LOD question gets at the technical skill that makes an artist employable.

How to ask the question without being a jerk

Some interviewers worry that asking about LODs is too technical, that it scares off artists who are "creative." That's nonsense. Every game artist should know the basics of the pipeline. It's like asking a chef about knife skills—it's not a trick question, it's a fundamental.

Frame it as a conversation, not an interrogation. Say something like: "I'm curious about your process. When you built this character, how did you approach the LOD chain? What were the triangle counts, and how did you decide what to cut?" Most artists with real experience will light up. They'll talk about the trade-offs, the silhouette, the normal maps. If they freeze, you have your answer.

You can also ask about a specific scenario: "You're building a character for a mobile game that needs to run at 60fps. The character has a complex cloak. How do you handle the LODs?" This tests not just knowledge but problem-solving. They should mention reducing the cloak's geometry, maybe using a simpler material, and balancing the budget across the whole character.

Bottom line

Stop asking about portfolios. Start asking about LODs and the high-poly-to-low-poly pipeline. It's the fastest way to find out if an artist can actually ship a game asset, not just render a pretty picture.

Sources

• Game art pipeline (Wikipedia) - https://en.wikipedia.org/wiki/Video_game_art_design
• Unity mobile optimization guide - https://unity.com/how-to/mobile-game-optimization-tips-part-1
• Technical artist roles (gamesindustry.biz) - https://www.gamesindustry.biz/technical-art-roles-what-they-are-and-how-to-get-one
• Unreal Engine documentation (Epic Games) - https://dev.epicgames.com/documentation/en-us/unreal-engine/nanite-virtualized-geometry-in-unreal-engine
• Level of detail (Wikipedia) - https://en.wikipedia.org/wiki/Level_of_detail_(computer_graphics)