“Can generative AI really speed up game production?”
That’s a question I found myself asking after reading an August 22 op-ed from Gamesindustry.biz features editor Lewis Packwood recounting his conversations with developers at Gamescom. According to Packwood, developers at the event were telling him about the need to make games on faster development cycles—targeting only one or two years of work instead of half a decade. Many developers we’ve heard from in the last year have echoed that sentiment, but the piece still proved controversial on social media.
Why? According to Packwood, developers are doing this by using generative AI to create code snippets, concept art, and other assets in order to “quickly iterate” on ideas. “AI is the games industry’s dirty little open secret,” Packwood wrote, saying many he spoke with at Gamescom were “coy” about the topic, dreading backlash like the kind faced by 11 Bit Studios after AI-generated assets were found in The Alters.
While Packwood cruised downriver to argue that because AI is “already in widespread use” in the game industry, then it’s “absurd” for developers to “live in fear forever,” I was still back upstream trying to understand how generating code snippets, concept art, or other assets could speed up production. The logic implies that what’s slowing down game development is the speed of making assets included in the game like code, 3D models, text, audio, etc. If that’s the problem, then generative AI spitting out assets would be a solution.
But is that the actual problem?
I decided to put it to the test with emails to developers, publishers, and industry experts at companies ranging from Tencent, to Hooded Horse, NEARstudios, and beyond—all asking one not-so-easy question: “what do you think is the number one reason games are taking longer to make?”
Their answers point in a different direction: one that centers the difficult-to-measure quality of games rather than the quantity of assets used to make them.
More assets doesn’t equal more fun
“The number one reason games are taking longer to make is that the quality bar has risen dramatically. Players today expect fun and engaging gameplay, cutting-edge visuals, memorable narrative and characters, and the ability to play whenever and wherever they want.”
That’s what Tencent Games Strategy and Compliance vice president Yongyi Zhu offered up in response to our hip-fired developer survey. So, if you’re a large studio making games in a familiar genre, it takes a lot of work to meet player expectations, let alone exceed them.
Virtuos managing director of game division Christophe Gandon echoed this sentiment. “As the games industry matures, projects grow in complexity as players increasingly demand better quality, immersion, and satisfaction in game length and value. Developers strive to meet or exceed these expectations.”
Zhu and Unity Engine senior vice president of product Adam Smith both pointed out that smaller teams making more experimental titles can move way faster. Smith nodded to Aggro Crab and Landfall’s viral climbing game Peak as a key example. “Many larger studios tie up resources in costly infrastructure before they can focus on making games, leading to higher costs and longer cycles,” said Smith.
The vicious cycle of photorealism
Smith’s point about “costly infrastructure” and the efficiencies of small teams made us eager to hear from NEARstudios CEO Heather Cerlan, a veteran technical artist from the world of large triple-A games. To her, the longer developer cycles are a consequence of tying the quality of games to photorealistic art direction.
“For the last 10 to 15 years games grew in scope and became more and more photorealistic. Hardware and tools were coming to market that technically made this feasible,” she said. “Developers were excited about the tech and were eager to use them and showcase the work made with those tools.”
But just because those tools made photorealism “feasible” didn’t mean it made them easy to use. Cerlan and her peers knew that. But plenty of “executives and decision-makers” didn’t… and they were the ones approving ever-skyrocketing budgets. Photorealistic games required complex tools, complex tools required specialists in numbers that ballooned team sizes. Bigger teams meant worse communication.
Worse communication meant unexpected tech problems that took time to address. Unexpected tech problems created performance optimization problems, and before you know it your game studio is spending time chasing frame rate bugs than iterating on “core gameplay” and other elements that “95 percent of players don’t care about.”
“[Then] when the game finally launches publicly, there’s frame rate drops and bugs that frustrate players and they lose confidence,” Cerlan concluded. “This problem is making it so that they couldn’t innovate in areas where players wanted innovation. The industry is now facing a reckoning, and I believe this issue of ‘more, bigger, prettier’ is a major factor as to why everyone in video games is suffering right now and why everything feels soulless.”
Image via NEARstudios.
Cerlan’s tool-oriented thinking casts a glaring light on the idea of better games emerging from a process where the asset machine is just cranked up to maximum speed. But photorealism isn’t the only reason a game may take years in development—what happens when developers report that spending more time in the shop is key to making a high quality game?
Can longer development cycles actually work?
Given that Hollow Knight: Silksong is riding high after 7 years of development we really should have been more prepared for developers to praise longer production timelines. Hooded Horse CEO Tim Bender and Riot Games senior game producer Patrick Miller flipped the script on us, each referencing the sky-high quality of player expectations for new games. “Players definitely have high expectations of quality and content for new games.
Extra time in development can fundamentally improve a game’s reception on release,” said Bender. “I couldn’t say how this factor weighs against all the others involved, but we are frequently in the position of encouraging developers to take all the time they need to get things fully ready before release.
“Even for an Early Access release that often means a lot of delay to ensure things are in great shape before anyone starts paying money for the game.”
Bender’s response echoes what Cerlan mentioned about players being let down by bugs and frame drops when playing triple-A games—but the difference might be about what kind of expectations you’re being measured against. Promise visual perfection, and every inadvertent T-pose damns you to the pits of Tartarus. Promise hours of emergent gameplay, and players will shrug off frame drops or hitches.
Meanwhile Miller said the nearly ten years of work on Riot Games’ League of Legends-themed fighting game 2XKO (previously known as “Project L”) was vital to make sure the game isn’t KO’d in round one. Calling himself “optimistic” on the topic of long development cycles, he said it’s important for developers to “take the time to better understand the game we should be making and the players we’re making it for.”
Miller said that in his many, many years working on 2XKO he saw the team go through “countless cycles” of building, testing, and refining the game’s direction. “Each time we do this, the team learns something important that shifts our direction slightly,” he said. “These longer cycles aren’t necessarily about having more room to fix bugs or build more content, they’re about giving the team more runway to solve problems at a reasonable pace instead of overloading them, so that the path to shipping and post-launch support are smoother.”
The stakes are all the more higher for 2XKO because it’s a fighting game—a genre defined by standout games with thirty years of mistakes and successes built into their DNA. “It takes a lot of time and effort to unpack the 30+ years of lessons that the veteran studios have learned,” said Miller.
“I often remind my teammates that it took seven Tekken games (and two Tekken Tag games) to get to Tekken 8. I’m glad we didn’t ship any of the ‘Project Ls’ we were making before we got to 2XKO.”
Generative AI’s princess is still in another castle
Can generative AI solve any of the problems described above? Probably not—at least not under Packwood’s framing. Even when new technology can streamline processes and improve efficiency, “the time saved is often reinvested in enhancing or creating more content,” Gandon explained to Game Developer.
The insights from Cerlan, Smith, and Zhu suggest that maybe smaller teams are the key to producing games at a faster clip—but that could lead to a world where “core” teams shrink to smaller and smaller proportions while underpaid contractors and external development studios fill in the gaps.
The solutions to extra-long development cycles don’t seem to be about tool choice or team size. Cerlan told us that she founded NEARStudios to “enable smaller teams to feel empowered to innovate, who could make strategic choices on fidelity and let go of those misinformed expectations that I believed were holding back video games.”
Scratch out the word “smaller,” and that thinking can work up and down the game industry if the right people are given the right shot.