NextFin News — Game Science has made clear that its second Black Myth game will not lean on generative artificial intelligence in the phases that shape its look and feel. In mid-August, Yang Qi, the studio’s co-founder and art director, answered a follower who asked whether current AI tools could accelerate Black Myth: Zhong Kui. His reply, posted on Weibo after he had praised Christopher Nolan’s new film The Odyssey, was direct:
“We’ll keep AIGC away from the design and asset stages. It’s not that we’re against it—we just want to hold the last shift properly. While we’re at it, look at digital cinema and 3D post-production: everyone says they’ve reached a dead end, yet Nolan is still going strong, stirring things up with older techniques. There’s no last train you absolutely have to catch. You can always check into a small inn for a while. And AI? There’s plenty of time for that. No rush.”
The comment stands apart from prevailing practice. Across the industry, generative systems are routinely used to produce concept art, textures, environmental assets and audio, trimming both calendar time and labor costs. The output is often functional yet frequently marked by generic detail and limited expressive range. Game Science is choosing, at least for the foundational visual work on its next title, to keep those decisions under direct human control.
Black Myth: Zhong Kui was announced at Gamescom in 2025 and remains in early development. Public materials are limited to a short cinematic trailer and a New Year promotional film. Studio descriptions point to a tonal shift from the high epic register of Black Myth: Wukong toward a more grounded, everyday atmosphere drawn from folklore surrounding the demon-queller Zhong Kui. Combat systems are expected to center on objects traditionally associated with the character. The visual ambition is higher-fidelity, photo-realistic environments and figures within a Chinese aesthetic framework, with narrative design adjusted for both domestic and international audiences. No release date has been set.
The restraint is easier to understand against the studio’s recent record. Black Myth: Wukong, launched in August 2024, became one of the most successful Chinese single-player action games ever released. Cumulative sales across platforms have been reported above 30 million copies, with more than half recorded outside China. Strong concurrent-player numbers on Steam, sustained critical reception and solid physical sales in several European markets established it as a genuine international event. After such a result, the incentive to accelerate a sequel is considerable. Game Science appears prepared to resist it.
Yang’s reference to Nolan supplies a useful parallel. In an era dominated by digital capture and extensive post-production, the director has continued to employ photochemical processes and large-format film for selected projects, arguing that older methods still yield distinctive results. The same logic is being applied to game art: established craft techniques need not be abandoned simply because faster alternatives exist.
For a series whose identity rests on dense, culturally specific imagery, the decision carries practical consequences. Generative tools generate volume efficiently; they are less consistent at the precise historical and stylistic detail that players associate with Black Myth. By retaining human oversight in the early design and asset phases, the team preserves the ability to set and maintain a coherent visual standard before any later production efficiencies are considered.
Many studios, Chinese and international, have publicly integrated generative pipelines to manage rising costs and tighter schedules. Game Science, buoyed by the commercial performance of its first major title and still early in the life of the next project, can afford a different tempo. Whether that slower approach yields a sharper artistic result, or simply a longer development cycle, will be clear only when the game reaches players. For the present, the studio has elected to treat visual craft as a resource worth protecting rather than a constraint to be automated.
Comments