Microsoft introduces an AI model for game development called Muse
Microsoft partnered with Teachable AI Experiences and the Xbox Games Studios" Ninja Theory to introduce a new AI model aimed at helping game developers. It"s called Muse.
What can Muse do? It can generate game visuals and controller actions based on a 3D understanding of a game. The model was trained on 1.6 billion parameters and with 1 billion images and controller actions. That"s about 7 years of continuous human gameplay. They used Ninja Theory"s Bleeding Edge game.
The data used for the model was collected using gameplay data from users who gave their consent. The model works with visuals and recorded controller actions at 300 x 180px resolution. As a result, Muse can generate sequences of up to 2 minutes and then predict game evolution in world model mode. The demos show the model successfully implements camera shifts, path choices and even entirely new characters. And it does so by respecting the game"s physics.
So, what"s Muse"s utility in today"s game development? Microsoft says it is already using Muse to develop real-time playable AI models trained in other games. It helps developers try out new concepts and maybe even expand on existing titles.
Another interesting application could be the easier implementation of retro games that work on old hardware. The AI model can be used to bring old classic games to modern machines.
Muse"s model weights, sample data and WHAM (World and Human Action Model) are all available on Azure AI Foundry. Short, interactive AI game demos will also be available via Copilot Labs sometime this year.
You can check out some of the demos at the source link below.
