Microsoft Research has unveiled the Muse Generative AI model, capable of generating game-like images that can be controlled by players in real-time, representing a current state of research. Microsoft sees the potential to develop and create new opportunities for game design.
Muse is a collaborative research project between Microsoft Research and Ninja Theory, a studio under Xbox, utilizing data from Ninja Theory’s game Bleeding Edge to train AI to understand the world of 3D games, game physics systems, and player responses through game controllers to create real-time games.
The latest Muse model, WHAM-1.6B, can generate gameplay based on specified prompts within the desired game, with results at a resolution of 300×180 pixels at 10 fps. It can also display fine details, such as interaction with in-game objects and movement in response to prompts.
Microsoft doesn’t intend for Muse to be an all-encompassing AI game creator but believes it can help innovate gameplay styles by learning from player data. It also sees the opportunity to revamp classic games without relying on old game engines but by using gameplay system studies and creating new environments in 3D.
Muse will soon be available for testing through Copilot Labs.
TLDR: Microsoft Research introduces Muse Generative AI for real-time game image generation and player-controlled experiences, aiming to revolutionize game design and create new gameplay opportunities.
Leave a Comment