Auctoria Uses Generative AI to Create 3D Video Game Models
-
Aleksander Caban, co-founder of Polish virtual reality game development company Carbon Studio, noticed years ago a significant problem in modern game design: manually creating basic elements in video game environments, such as rocks, hills, paths, etc., is often a time-consuming and tedious process. To address this issue, Caban decided to develop technology to automate this process.
He collaborated with Michal Bugała, Joanna Zając, and Carbon Studio's other two co-founders, Karolina Koszuta and Błażej Szaflik, to launch the Auctoria platform, which uses AI to generate 3D video game assets from scratch.
Zając stated in an email interview: "We created Auctoria out of a passion for limitless creativity. Its original purpose was to support the work of game development professionals, but anyone who wants to create can use it. Professionals lack advanced tools—most tools are aimed at hobbyists and amateurs. We wanted to change that."
Auctoria leverages generative AI technology to produce various types of video game models. One feature of the platform is its attempt to generate complete 3D game levels, including paths for players to explore (albeit relatively basic ones). Another feature converts uploaded images and textures of walls, floors, and pillars into corresponding 3D models. Users can also input text prompts to have Auctoria generate assets, similar to DALL-E 2 or Midjourney. Alternatively, they can provide a sketch, and the platform will attempt to transform it into a usable digital model.
Zając claims that all the AI algorithms powering Auctoria, as well as the data used to train them, were developed in-house. "Auctoria is entirely based on our content, so we don't rely on any other vendors," she said. "It's a standalone tool, independent of any external engines or open-source solutions."
However, in the market for AI tools that generate game assets, besides Auctoria, there are 3D model creation platforms like 3DFY and Scenario, as well as startups such as Kaedim, Mirage, and Hypothetic. Even established companies like Nvidia and Autodesk have begun venturing into this field, launching applications like Get3D (which converts images into 3D models) and ClipForge (which generates models based on text descriptions).
Meta has also experimented with using prompts to generate 3D assets. Last year, OpenAI released Point-E, an AI that synthesizes 3D models with potential applications in 3D printing, game design, and animation.
According to Proficient Market Insights, the 3D model market could be worth $3.57 billion by 2028.
But Zając notes that Auctoria's relatively long development cycle—approximately two years of R&D—has given it a more robust and comprehensive toolset compared to some competitors. "Currently, there's a lack of AI-based software capable of creating complete 3D world models," Zając said. "Existing solutions typically consist of 3D editors and plugins, but they only offer a fraction of Auctoria's capabilities. Our team started developing this tool two years ago, enabling us to have a ready-to-use product."