An Unexpected AI Innovation is Changing the Landscape of Game Development

The video game Valorant, a dynamic team-oriented shooter, has recently emerged as a proving ground for an exciting new direction in artificial intelligence research. The game’s creators at Riot Games (a subsidiary of Tencent) are implementing 3D-native AI models to prototype new characters, environments, and narratives, according to a researcher familiar with the company’s initiatives who spoke on condition of anonymity.
While many AI models excel at generating text, images, and video, Tencent’s Hunyuan (混元 or “first mix”) family of models can conceive 3D objects and interactive settings. The source indicates that Tencent’s models are also utilized by the developers of another Tencent game, GKART, along with some independent developers. Tencent has chosen not to comment.
“The games industry demands significant investment,” the source remarks. “Previously, you would require a month to design a character. Now, you can simply input text, and Hunyuan can present you with four options in just 60 seconds.”
This news serves as an initial indication that models proficient in understanding and recreating the physical world could become standard in game design. Beyond generating game content, these models have the potential to advance virtual and augmented reality and assist robots in acquiring new skills.
“There’s a significant surge in 3D vision research at the moment,” states Alexander Raistrick, a graduate student at Princeton University who is exploring innovative methods for generating 3D content. “There are numerous high-impact applications: content creation, self-driving technologies, and a range of challenges related to augmented reality.”
Raistrick notes that video games are a clear application for 3D AI models. “Producing 3D meshes [a common method for representing 3D objects] is typically a foundational aspect of game development,” he explains.
However, as seen in other creative domains, the use of AI in video game development raises controversy. Concerns about job displacement due to AI are substantial. Some developers argue that games should clearly indicate when they feature AI-generated content. Others believe it may be too late: The technology is already widespread within the industry.
In July, Tencent launched HunyuanWorld 1.0, a model designed to generate interactive scenes. I tested it a few months ago, navigating through a scene reminiscent of a Lego movie—a valley filled with vividly colored blocks extending into the horizon. More recently, I experimented with a simpler model, Hunyuan 3D, which can create 3D objects. I used it to generate appealing custom Dungeons & Dragons characters for 3D printing. In October, Tencent unveiled an updated version of HunyuanWorld that allows users to upload videos for generating 3D scenes.
