The deck looked good. Twelve slides of character renders, three environment flythroughs, a logo wall with two names you recognized. Six weeks into production, you are on a call asking why the third asset batch came back with silhouettes drifting and texel density inconsistent across a single environment set.
Nobody signs a weak art vendor on purpose. The problem is that the evaluation most studios run — look at the portfolio, check the rate card, call a reference—tests almost none of the things that predict whether production holds together. Here is what does.
The portfolio is the least useful thing in the pitch
A portfolio proves a studio can produce a finished image. It does not tell you who produced it, whether that person still works there, how many revision rounds it took, or whether the piece ever shipped in an engine.
That last one matters more than people expect. Concept art that never had to survive a technical constraint is a different discipline from production art that had to hit a polygon budget and a memory ceiling on a five-year-old handset. Both look fine in a PDF. Only one of them has been tested.
Ask which pieces shipped, and what the shipped version looked like next to the portfolio version. A studio with real production history will have an answer and will usually volunteer the difference before you ask. The shipped asset is almost always the less pretty one, and they will tell you exactly why.
Art is not a line item you can treat as marginal. Industry cost breakdowns for 2025–2026 consistently show production—art chief among it—running 40–60% of total budget, depending on the visual direction chosen. You are not buying pictures. You are buying the process that produces them at volume.
Style range is not the same as style depth
A portfolio showing eight distinct visual styles reads as versatility. Sometimes it is. More often it means eight different contractors passed through, and the one who made the style you liked left eighteen months ago.
Depth looks different. It looks like twelve assets in one visual language that hold together as a set — same lighting logic, same material treatment, same level of abstraction in the silhouettes. That is much harder to fake and far more predictive of what arrives at volume.
If the style you want appears exactly once in their portfolio, ask who made it and whether that person is still on the team. It is a blunt question. Good studios answer it directly.
There is a related trap with scale. A team that delivered forty assets in a consistent style has proven something a team that delivered four has not. Ask for the largest set they have shipped in the direction you are after, and ask what broke at that scale.
Ask how one asset actually got made
Good game art services are built around a repeatable process, not a single flourish — and that process is what you’re actually testing when you evaluate a vendor. Pick a single asset from their portfolio and ask them to walk you through its production history. Not the style—the schedule.
How long from brief to first blockout? How many review gates? Who signed off at each one? What came back for rework, and was that rework a craft problem or a brief problem?
The answers tell you three things a rate card never will. Whether they work to a defined pipeline or improvise per project. Where their internal review sits, because studios that only review at delivery will hand you problems you then have to catch yourself. And how they handle a brief that turns out to be wrong halfway through, which is the normal case rather than the exception.
Studios buying art this way on a fixed-bid basis often find the bid covered production but not the rework cycle, and the second invoice is where that surfaces. Ask what triggers a change order before you sign, not after.
One more question worth asking: what is the smallest change that would require a rebuild rather than an edit? If they can answer precisely, they have thought about your pipeline. If they hedge, they have not.
Revisions are where budgets actually die
Most art overruns are not caused by weak art. They are caused by a revision structure nobody agreed on in writing.
Settle two numbers before signing: how many revision rounds are included per asset, and what counts as a round. A studio that includes three rounds but treats a batch of eleven notes as three rounds has given you one round with extra steps.
Then settle what happens when the brief changes. If your creative director redirects the visual language in week nine—and this happens—is that a revision or new scope? The contract should say. A verbal understanding reached at kickoff will not survive the conversation you have in week nine.
Production schedules commonly reserve a 20–30% overflow buffer on every milestone, a rule of thumb across the industry, and the honest price of uncertainty.
Art is usually where that buffer gets spent first. The criteria that separate an AAA game art studio from a competent mid-tier one show up here more clearly than in any portfolio, because at AAA asset volume an undefined revision policy stops being an annoyance and becomes a schedule risk.
What the Week Six Call Is Really Telling You
Go back to that call about drifting silhouettes and inconsistent texel density. Almost none of it was an art skill problem. It was a process problem that already existed at signing and only became visible under volume.
Which is the useful reframe. You are not evaluating whether a studio can make good art. Most of the ones that reach a pitch can. You are evaluating whether their process survives contact with your schedule, your engine constraints, and a creative director who changes their mind in month three.
The portfolio answers the first question. Only the process questions answer the second, and those are the ones most evaluations skip.
Frequently Asked Questions
How much should a first-time studio budget for outsourced game art?
Scope drives this more than style. As a reference point, coverage of GDC 2026 reported that half of indie studios outsource art at an average spend near $20,000, which buys a defined asset set rather than a full production. Larger 3D pipelines with cinematic assets run past $500,000. Count your assets before requesting a quote, because vendors price volume, not ambition.
What is the minimum information a vendor needs to quote accurately?
Asset count by type, target engine and platform, polygon and texture budgets, and a visual reference set. Without technical constraints attached, any quote you receive is a guess that will be revised later. Studios that quote confidently from a mood board alone are pricing the attractive version, and the number moves once real limits arrive.
Is a paid pilot worth it before committing to production?
Usually yes. A paid test on two or three real assets costs a fraction of a production commitment and tests the one thing a portfolio cannot show you: how a team responds to your notes. Watch their reaction to round two more closely than the first delivery. Working style shows up there.
