Tencent’s new AI can create open-world video games from text

Tencent’s GameGen-O has raised the stakes for AI-generated video games, enabling the generation of limited open-world games from text.

This procedurally generated technology aims to dramatically reduce the time and cost of building open-world games and the content that players interact with. Tencent has worked with students in Hong Kong to develop this dataset of gaming content, which can independently generate environments, characters, and gaming assets.

The emergence of this technology could result in lower overheads for gaming studios, which have been slimming down the workforce in an attempt to save money over the past few years.

The research team on GameGen-o believes “It underscores the potential of generative models to serve as an alternative to rendering techniques, which can efficiently combine creative generation with interactive capabilities..”

Tencent presents GameGen-O

Open-world Video Game Generation

We introduce GameGen-O, the first diffusion transformer model tailored for the generation of open-world video games. This model facilitates high-quality, open-domain generation by simulating a wide array of game engine… pic.twitter.com/DlBt9iiLYZ

— AK (@_akhaliq) September 13, 2024

Artists and developers spend months creating our favorite games’ characters, environments, actions, and events. Tencent, the Chinese giant behind this new AI software, has a diverse portfolio with stakes in global gaming companies. Valorant’s Riot Games, Ubisoft Entertainment, Baldur’s Gate 3’s Larian Studios and Remedy Entertainment.

The AI’s researchers have named the process for creating these environments the open-world Video Game Dataset (OGameData). This bank of gaming information reportedly includes 150 next-gen games and 4,000 hours of video clips, according to Geeky Gadgets.

The game software takes a two-stage process to develop the content, according to the developers of GameGen-O, are “foundation model pretraining and instruction tuning. In the first phase, the model is pre-trained on the OGameData via the text-to-video and video continuation, endowing GameGen-O with the capability for open-domain video game generation. In the second phase, the pre-trained model is frozen, and we fine-tuned using a trainable InstructNet, which enables the production of subsequent frames based on multimodal structural instructions.”

The process is still in early stages, but characters like The Witcher’s Geralt of Rivia, Red Dead Redemption’s John Marston and Ghost of Tsushima’s Jin Sakai have all been rendered in the OGameData according to the GitHub announcement page.

As we reported, other companies, such as Google, have made advancements in AI-generated content. The world’s largest search company released a paper that went into great detail about GameNGen and its process for creating Doom-like gameplay on a neural network.

Image: Midjourney

The post Tencent’s new AI can create open-world video games from text appeared first on ReadWrite.