Square Enix-Backed Atlas Uses AI to Turn Text and Images Into Game Worlds

A Vienna-based 3D generative AI platform emerged Tuesday on the red-hot AI game development scene flush with $6 million in funding. In shedding its stealth status today, Atlas says it will develop better AI models for the gaming industry, where it is already shortening the typically long development process to a few prompts and a few hours of AI generation—putting the industry within reach of many more creators.

“One of the reasons we’re coming out of stealth now is we’re going to launch an alpha version of the platform later this year,” Atlas CEO Ben James told Decrypt. “We want to enable smaller, medium, and indie game developers to leverage [Atlas] to be able to build out their own worlds and facilitate some of these more impossible experiences.”

Launched in 2020, Atlas offers a 3D AI platform that aims to accelerate the creation of virtual worlds and experiences for game developers and brands building video games and extended reality (XR) environments.

“In recent years, the rising cost of game development has been challenging for the entire game industry, and gen AI is expected to help streamline the process in new and exciting ways,” Hideaki Uehara, Square Enix General Manager of Investment & Business Development, said in a statement. “We are excited about Atlas’ unique technology and look forward to seeing how it might unlock efficiencies in our business.”

The $6 million came in two funding rounds, a $4.5 million round led by 6th Man Ventures and a $1.5 million round led by Collab+Currency. Other investors joining the fundraise include Final Fantasy and Kingdom Hearts game developer Square Enix, the a16z Scouting Fund via Shrapnel, Contango, Gaingels, GFR Fund, New Renaissance Ventures, Wagmi Ventures, Flamingo DAO, Founders Inc., Landvault, Monaverse, Neon DAO, and Slope Fund.

Thanks to recent developments in image-generating artificial intelligence like Midjourney, Stable Diffusion, and Runway, game developers are turning to AI to design characters and landscapes that would have taken months, even years, to create in the past. Indie studios could conceivably better compete with larger AAA game studios.

There are already some big players in the space. In May, Nvidia unveiled its Avatar Cloud Engine (ACE), which uses generative AI to create game scenes and non-player characters. In August, Sweden-based Hiber launched Hiber3D, which uses Google’s generative AI technology in its online game development platform, Hiberworld. Last month, Nvidia joined forces with 3D software publisher Masterpiece Studio to release Masterpiece X, designed to make 3D modeling as easy as creating a two-dimensional image.

But while generative AI can give an indie studio an edge, James acknowledges that the technology is a double-edged sword.

“If you’re a bigger studio, you tend to focus on the time and cost savings that are applicable from this type of technology versus if you’re a smaller studio, you’re probably interested in those games and experiences that you haven’t thought about before,” James said. “But I would say is that our tech aims to just supplement a designer’s creative intuition.”

James emphasized the importance of building AI technology on ethically sourced training data, prioritizing privacy, and protecting intellectual property. Atlas was trained on its own data and not using datasets from other game developers using sparse representation, he explained, a method of encoding information where only a small number of elements are used instead of many.

“It’s helpful because it means that we can close off these models in a way that if your specific IP is going to produce a particular AI generator for your game that really becomes your AI model that nobody else can use,” James said. “Nobody would even be able to generate the same results you would unless they had your your underlying IP as the input into that process.”

As part of a research grant, James added that the Atlas platform was also built following the European Union’s Ethical Guidelines for Trustworthy Artificial Intelligence, which elevates the principles of transparency, privacy, human agency, oversight, and inclusion.

“I think it’s important to build up with this kind of principles in mind because we see it all over the place how important this becomes, especially in an industry like gaming where IP is so critical,” James said. “You really want to work with trustworthy applications.”

Looking to the future, James said he was optimistic that Atlas would evolve to include a broader range of AI-powered tools that simplify complex tasks like 3D modeling and coding.

“We’re working with fantastic partners, but we’d like to start to expand and grow,” James said.

Edited by Ryan Ozawa.