Everything a production lead needs to brief, budget, and approve character art: the stages, the deliverables, the 2026 cost picture, and where AI actually helps.
Developers pushed a record 21,335 games onto Steam in 2025, up from 18,475 the year before, according to SteamDB. Of the 19,112 that had shipped by December 12, 9,327 never reached ten reviews. Most players never see a store page at all. They see a 462 by 174 pixel capsule in a search result, and at that size a game is a silhouette and two colors. Everyone in this industry knows that, and character art is still the line item that gets cut first. Mobile is the same problem in a tighter frame: an icon, a screenshot carousel, and an ad creative fighting for a thumb in a feed. Character concept art is where that fight is won or lost, long before a single line of gameplay code is written.
That pressure is arriving just as the art labor market moves the other way. The US Bureau of Labor Statistics projects 2 percent employment growth for special effects artists and animators between 2024 and 2034, against 3 percent across all occupations, on a base of 57,100 jobs and a median wage of $99,800 as of May 2024. More games, more art per game, and a hiring pool that isn’t expanding to match. Then add generative AI: GDC’s 2026 State of the Game Industry found 36 percent of professionals using the tools, while 64 percent of visual and technical artists said the technology is hurting the industry, the most negative reading of any discipline surveyed.
Nobody in production asks “is this art good” any more. They ask who makes it, by when, for what, and what exactly lands in the modeler’s folder on the last day. We’ve been on the receiving end of that question for eleven years, and the fourth part is the one that costs money when nobody answers it. This guide covers the creation stages and what each one produces, how to choose between an in-house hire and an outsourcing partner, what the work costs in 2026, how to read a portfolio before you sign anything, and where AI belongs in the pipeline.
Character concept art is the visual design work that turns a written character description into a specific, buildable image: proportions, silhouette, palette, costume logic, expression range, and the small details that tell a player who this person is before they say a word. It exists so that the writer, the art director, the animator, and the 3D artist are all looking at the same character instead of five private versions of one paragraph.
That last part is the whole job. Two people can read “grizzled bounty hunter, mid-forties, ex-military” and picture entirely different men. The concept artist’s output is the agreement.
David Cage, in his TED talk on how video games turn players into storytellers, makes the argument for why this matters commercially: a character the player can shape turns that player into a co-author, and co-authors come back. A character nobody feels anything about is a model with an animation set attached.
What actually lands in your drive at the end of the process is less romantic than the phrase suggests:

A concept that looks stunning as a single hero render but has no turnaround and no material notes is a poster. The modeler will come back with questions, and those questions cost more than the second view would have. If you’d rather hand the whole package to a team that already works this way, our game concept art services page covers scope, deliverables, and engagement models.
You already know the outcomes worth aiming at. Darth Vader, Steve from Minecraft, Mario: readable at any size, imitable by a child with a marker, instantly separable from every other character in their own game. What’s less obvious is that all three came out of a process with recognizable stages, and that skipping one shows up later as rework, never as a missing feature.
The formula behind a character that sticks breaks into five parts:
Here is how those parts map onto a production schedule and, more importantly, who has to sign off before the next stage can start.
| Stage | What it produces | Who signs off | Typical duration |
|---|---|---|---|
| Brief and reference gathering | Written character brief, mood board, style targets | Art director, narrative lead | 2 to 4 days |
| Thumbnails and silhouettes | 20 to 40 rough options, 3 shortlisted | Art director | 3 to 5 days |
| Refinement | One chosen direction rendered with material and costume callouts | Art director, creative director | 5 to 8 days |
| Turnaround and sheets | Turnaround, expression studies, prop breakouts | Lead 3D artist | 4 to 6 days |
| Handoff and support | Layered source files, style guide entry, revision window | Production lead | Ongoing through modeling |
Source: Innovecs Games delivery data, projects staffed 2023 to 2026.
Read the sign-off column before the duration column. Most schedule slips we see aren’t caused by the drawing taking longer than expected; they’re caused by a stage sitting in a reviewer’s inbox because nobody agreed in advance who owned the approval.
Every duration in the table above assumes the artist got a usable brief on day one. Most don’t. Three adjectives and a mood board is a wish dressed as a brief, and the first week disappears into an interview that should have been a document.
A brief that works fits on one page and answers six things:

One thing doesn’t belong: art direction written out as prose. If your description is detailed enough that only one answer is possible, you’ve drawn it yourself in words, and you’ll get back exactly what you specified instead of something better.
A good studio will interview the missing parts out of you. That still costs a week, and it’s a week nobody put in the plan.
Video game character concept art has to survive an engine constraint, an animation rig, and a store thumbnail. Almost everything that goes wrong on those three fronts is already visible at sketch stage, where fixing it costs an afternoon. Worth running the list below before the character goes into the game design document:
The test an experienced artist applies is simple: hand the sheet to someone who has never read the brief and see whether they can describe the character back to you. Posture, silhouette, clothing, palette, and accents should carry that on their own. When video game character concept art needs a written explanation to make sense, the design isn’t finished yet.
The 3D characters that come back wrong are almost never drawing failures. In our experience they’re a producer and an artist who both said “heavy” and meant two different silhouettes by it. Skill and toolset matter. Shared vocabulary matters more, and it’s the thing nobody puts on the checklist.
That’s why 3D character concept art needs more people in the room than 2D does. The concept artist, the 3D artist, the animator, and the technical artist all constrain each other, and style choice alone (photorealistic, semi-realistic, or stylized in the Pixar direction) moves topology budgets, texture memory, and how much the face can emote.
The build pipeline that follows the concept stage runs in a fixed order:

Animation comes after, and it’s where the character’s personality either shows up or doesn’t. Retopology and rigging are where art hands the character over to engineering, which is why studios often staff that boundary through a game technical art outsourcing studio rather than asking a concept artist to solve deformation problems they were never trained for.
Concept art is spiky work. A studio needs enormous art capacity for eight weeks of pre-production, then a fraction of it for the following year, then another spike when the roadmap adds a season pass. Hiring a permanent artist against a spiky curve means paying for troughs; staffing the whole thing externally means paying a premium for someone else to carry that risk. Neither is wrong. The choice usually comes down to how many characters you’ll need this year and whether the style is settled.
| Model | Ramp-up | Cost shape | Style range | Pipeline ownership | Honest downside |
|---|---|---|---|---|---|
| In-house concept artist | 2 to 4 months to hire and onboard | Fixed salary plus overhead | Narrow, whatever that artist is strong at | Full, stays with you | You pay through the troughs, and one artist is one style |
| Freelancer | Days | Per asset or per hour | Wide, if you keep re-sourcing | None, you carry it | Availability disappears the moment a bigger client calls |
| Outsourcing studio or dedicated art team | 1 to 3 weeks | Monthly rate or scoped package | Wide, multiple artists behind one art lead | Shared, documented at handoff | Costs more per month than a freelancer, and the first two weeks go into learning your style guide |
Most disagreements we have with a client’s plan happen on the boundary between two of those rows, not inside one of them. A studio that needs four characters this quarter and none next quarter is looking at a freelancer or a scoped package, whatever the org chart says it wants.
The decision usually announces itself. Look for these:
Concept art outsourcing buys a functioning art production pipeline. The images are the visible part; the pipeline is what you’re actually paying for.
Six things belong in the contract, and all six are fair to insist on:

Vetting is the part teams rush. A paid trial on one character costs a few hundred dollars and a week; discovering the same information in month three costs the milestone. Ask for a paid trial project on a real character from your actual brief, not a portfolio review alone, and judge the studio on how it handles round two of feedback. The first delivery tells you very little. Rates vary far more than most buyers expect, from roughly $200 for a single character sketch to $25,000 and up for a AAA visual development package, and we’ve broken the drivers down in detail in our guide to concept art creation costs.
Co-development is the third option, and it’s where most of our mid-size accounts have ended up within a year of starting: external artists inside your pipeline, your tools, your review cadence, carrying part of the milestone rather than delivering to a spec and leaving. It also asks more of you than the other two, because shared responsibility means shared decisions.
If your roadmap has characters on it and no artist against them, a dedicated art team can be reading your style guide inside two weeks. Most of the calls we take start about a month later than they should have. Send us the brief and the milestone date, and we’ll scope the roster, the rounds, and the handoff format.
Players who connect with a character keep playing, buy upgrades for that character, and start modifying the hero’s look and posting the result. That’s the mechanism, and it’s the reason character art survives budget cuts that kill other art.
Which raises the harder question. Everyone in this market can draw. Telling apart the artist who can draw from the artist who can solve your problem takes about twenty minutes if you know what to ask for, and the industry’s default use of those twenty minutes is scrolling finished renders.
The baseline is easy to state and hard to find in combination:
That last one rarely shows up in a portfolio and never shows up in a rate card. It shows up in week two, when the first round of feedback either lands or doesn’t.
A portfolio shows finished work, which is the least useful thing to look at. What predicts performance is process evidence. Whether you’re evaluating a studio or a single concept artist for hire, ask for these:
Nobody trains producers to review art, and the industry has been fine with that for thirty years. The default is to review against taste, which produces notes nobody can act on and rounds that never converge.
Review against the brief instead. Four checks take ten minutes and catch most of what actually goes wrong:

Then write feedback that names the problem instead of the fix. “The silhouette reads like the merchant NPC” gives an artist somewhere to go. “Make the shoulders bigger” hands over a solution and closes off the four better ones the artist would have found.
Two process rules save more time than all four checks combined. Consolidate every reviewer’s notes into one voice before they go out, because contradictory feedback from three people burns a full round and a lot of goodwill. And agree in writing what counts as a revision round before work starts, not when you’re already on round four.
Taste notes are legitimate, by the way. Label them as taste. An artist can absorb “this is a preference, not a failure” without hearing it as a verdict on their craft.
Pixar register, Minecraft-inspired blocks, full photorealism: it all starts with drawing, detailing, backstory, and iteration, and the visual decisions are the same ones about posture, outfit, accessories, and silhouette. What changes downstream is only the cost of changing your mind. A silhouette revision at thumbnail stage costs an afternoon. The same revision after rigging costs a sprint and a renegotiated milestone.
That’s the entire commercial case for pre-production. Concept art isn’t expensive. Concept art decided late is.
Players downloaded games 52 billion times in 2025, and Sensor Tower’s State of Gaming 2026 recorded a third consecutive year of in-app purchase revenue growth across the App Store and Google Play. Read the growth rate instead of the total, though, and the picture sharpens. Gaming IAP approached $82 billion in 2025 but added only 1.3 percent year over year, while non-game apps put on 21 percent. The easy growth is gone, which leaves unit economics, and in a mobile game the character is frequently the product being sold. Skins, seasonal variants, and cosmetic bundles all monetize a design that players already care about, and none of them work on a character nobody recognizes.
Recognition has a technical component that concept art controls directly. A character has to survive a kill feed, a crowded end-of-match screen, and a shop tile the player scrolls past in half a second. That’s a silhouette and palette problem decided in week two of pre-production, and there’s no marketing budget that fixes it afterward. Player retention follows the same logic: people return to a world they can picture, and they picture it through its characters.
In mobile, the character does a second job that console teams rarely have to think about: it carries the ad. Unity’s guidance on reading in-ad metrics points at three numbers, engagement rate, completion rate, and click-to-store rate, and all three begin with whether a player can tell what they are looking at inside the first second of a feed.
That makes concept art a user acquisition asset, not just a production one. A design with a clean silhouette and a two-color read survives compression, a small screen, and a thumb moving at speed. One that only works as a full hero render doesn’t, and the CPI tells you so inside a week.

It also changes what you ask the artist for. Alongside the turnaround, ask for a variation pack: the same character in three poses, two palettes, and one exaggerated expression, all on-model. That is the raw material for a creative test, and it costs a fraction of commissioning three separate characters to find out which one converts.
One caution, because this argument gets oversold. Better art lowers CPI on a hook that already works. It does not rescue a bad one, and no art budget has ever fixed a game nobody wanted to play.
The data here is more interesting than the discourse. GDC’s 2026 survey found 36 percent of industry professionals using generative AI, and Game Developer’s breakdown of the same report shows what they use it for: 81 percent for research or brainstorming, 47 percent for daily tasks like email and code assistance, 19 percent for asset generation, and 5 percent for player-facing features. Adoption is clustered at the ideation end of the pipeline, which is exactly where concept art lives. That’s either reassuring or alarming depending on which end of the pipeline you sit at. Meanwhile 52 percent of respondents said the technology is having a negative impact, up from 30 percent a year earlier, and only 7 percent called it positive.
Here’s the productivity claim with the marketing taken off: AI compresses the widest and cheapest part of the process, and leaves the expensive part alone. That’s still worth having. It just isn’t what the pitch decks say. Where it earns its place:

AI image generation tools have not compressed the parts that consume most of a schedule: turnarounds that stay on-model across four views, sheets a rigger can build from, and revisions that respond to a specific note rather than regenerating something adjacent.
Two constraints keep human authorship in the middle of this process, and neither is about quality.
The first is legal. The US Copyright Office’s report on copyright and artificial intelligence, Part 2 of which landed in January 2025, concluded that purely AI-generated output isn’t protectable and that prompts alone don’t establish authorship, while human-authored contributions to an AI-assisted work remain protectable. For a studio, that means the ownership of a flagship character depends on a documented human contribution. Platform holders are asking about it too: Valve’s Steamworks content survey requires developers to disclose pre-generated and live-generated AI content and to confirm the game contains no infringing material. Both of those belong in your contract with any art partner, and they belong there before the first asset arrives.
The second is judgment. Deciding which of forty thumbnails carries the narrative, knowing that a costume detail will fight the rig, choosing a palette that separates a character from a level you haven’t built yet: none of that is a generation problem. It’s a decision made against context the model doesn’t have.
Our position, for what it’s worth: the price of a first draft has gone to nearly zero, and the price of judgment hasn’t moved at all. That’s why 64 percent of visual and technical artists in the GDC survey read the current moment as a threat while their own studios keep hiring art leads. The work didn’t disappear. It moved up a level.
Innovecs Games has spent 11 years in game development, shipped work on more than 300 games for over 100 clients, and staffs 200-plus developers and artists, including full-time concept specialists. Jam City, Gameloft, and 7 Elements Studios are among the studios we’ve worked with. That history matters here for one reason: character work that has to match an existing art bible behaves very differently from character work on a blank page, and we’ve done a lot of both.
We work three ways, and the differences matter more than the names. A dedicated art team is right when the roadmap has continuous character output; it’s the wrong answer for four characters and a gap. Project-scoped delivery fits a fixed brief and asset list, a launch roster or a seasonal drop, and it’s the model we most often start on. Co-development puts our artists in your pipeline, your review cadence, and your milestone structure, and it’s the one that asks the most of both sides: your leads have to direct people who don’t sit in your building.
Handoff is the part worth checking before you sign anything, ours included: layered source files with named groups, turnarounds, model sheets, and a style guide entry written for whoever inherits the character after us. We assume that person won’t be us. Concept art outsourcing works best when the partner also covers what happens downstream, which is why the same teams extend into our broader game art services when a project needs environments, props, or UI in the same register.
If you need a single concept artist for hire rather than a team, we scope that too. Smaller engagements are how most of our long-running accounts started.
A character concept artist designs how a game’s characters look before anyone models or animates them. The role covers research, thumbnail sketching, silhouette exploration, costume and prop design, palette decisions, and the turnarounds and sheets the 3D team builds from. It sits between narrative and production: the artist translates a written description into a specific, buildable design. Most work in Photoshop or Procreate, and the stronger ones also handle ZBrush or Blender well enough to talk to modelers in their own terms.
Start with the brief and reference gathering, then move to twenty or more thumbnail sketches focused on silhouette rather than detail. Shortlist two or three, refine one into a rendered front view with material and costume callouts, then produce the turnaround, expression studies, and model sheet. Handoff includes layered source files and a style guide entry. Budget roughly two to four weeks for a hero character, less for supporting cast reusing an established style.
It’s a studio convention rather than a published standard, and you’ll hear it stated several ways. The common version: roughly 70 percent of a design carries the dominant shape language, material, or color, and the remaining 30 percent provides contrast. Applied to shape, a character reads as mostly angular with rounded accents, or the reverse. Applied to palette, one color dominates while a second creates the focal point, often at the face or the weapon. The point is hierarchy. Designs that split evenly between two ideas read as indecisive at a distance, which is exactly where recognition happens.
Freelance rates run from about $25 to $500 an hour depending on experience and complexity. Per asset, a single character typically lands between $200 and $1,500, highly detailed work between $500 and $5,000, and AAA visual development packages from $25,000 upward. Studio engagements are usually quoted monthly or as a scoped package, not by the hour. Those ranges and what moves them are broken down in our guide to concept art creation costs. The biggest cost drivers are the number of revision rounds, whether a turnaround and sheets are included, and whether commercial licensing and full IP assignment are in scope.
2D concept art defines the character’s appearance: silhouette, costume, palette, expression, and the design language that makes them recognizable. 3D concept work carries the same decisions into a form the pipeline can build, which means thinking about topology, deformation, texture budgets, and how the face will emote once rigged. In practice 3D is more collaborative, because modelers, animators, and technical artists all constrain the design. Almost every 3D character still begins as 2D thumbnails, since exploring shape is far cheaper on a flat canvas.
Volume and style stability decide it. If you need continuous character output all year and your style is settled, an in-house artist is cheaper per asset and builds institutional knowledge. If your need is spiky, or the style isn’t settled, or you need several registers explored quickly, outsourcing gives you range without carrying the troughs. A realistic in-house hire takes two to four months to source and onboard; an external team can be inside your style guide in one to three weeks. Many mid-size studios run both, with an internal art lead directing external artists.
Photoshop remains the default for painting and layered handoff, with Procreate common for sketching and thumbnails on tablet. ZBrush handles sculpting and high-detail exploration, Blender and Maya cover 3D blocking and turnarounds, and Substance Painter handles texturing further down the pipeline. AI image generation tools now appear at the reference and ideation stage in many studios. Tool lists matter less than whether the artist’s files arrive in a form your modelers can open and navigate without a phone call.
A hero character with full turnaround and sheets typically runs two to four weeks from brief to handoff, assuming feedback arrives on schedule. Supporting characters in an established style take a few days to a week each. Drawing speed is almost never what moves the date. Approval latency is. Name the approver for each stage before work begins, and most of your schedule risk goes away for free.
For internal use, frequently. For shipped assets, rarely without substantial human work on top. GDC’s 2026 survey found only 19 percent of AI users applying it to asset generation, against 81 percent for research and brainstorming, which reflects where the output actually holds up. Ask a current tool for the same character seen from behind and you will frequently get a different character. Add the distance between a striking image and a sheet somebody can actually build from, and that’s where the saved hours go back in. There’s also the ownership question: the US Copyright Office has held that purely AI-generated material isn’t protectable, so a flagship character built that way carries real IP risk.
Mostly at the front. Reference gathering and mood boards compress from days to hours, and artists can explore more silhouette variations before committing to a direction. Iteration on an approved design (palette variants, costume alternatives, lighting studies) also moves faster. What hasn’t changed is the expensive part: judgment about which direction serves the narrative, whether a design will survive the rig, and how a character reads against levels that don’t exist yet. Studios that get value from these tools tend to use them to widen the exploration phase, not to shorten the decision one.
Tell us the character count, the style target, and the milestone date. We’ll come back with a roster, a revision structure, and a delivery schedule, and we’ll say so if character concept art turns out not to be your real bottleneck. That conversation takes about a week.