Microsoft has announced Muse, a generative AI model that generates gameplay sequences by forecasting visuals and controller inputs. Created in partnership with Xbox Games Studios’ Ninja Theory, Muse is not an AI that plays games in real time but a tool that assists developers in visualizing and experimenting with new gameplay concepts.
Muse can generate extended sequences based on just one second of real human gameplay
In the research report shared by Microsoft, it has been showcased that by examining only one second of actual human gameplay, Muse can produce entire sequences that appear and feel like the original game. This AI is a part of Microsoft’s World and Human Action Model (WHAM) research and was trained on Bleeding Edge, a multiplayer game by Ninja Theory, through gameplay data.
Muse can analyze how a game is likely to unfold based on player moves and provide developers with new avenues to experiment with mechanics and level design. Microsoft proposes it could also be used to save old games by making them compatible with modern hardware.
To enable additional research, Microsoft has released Muse’s model weights, sample data, and a prototype tool named the WHAM Demonstrator on Azure AI Foundry. The tool enables developers to engage with the AI, modify gameplay sequences, and experiment with how AI-generated content might be used in game development.
Though Muse is a significant advance in AI-powered game development, it’s an early experiment. It can create great moments of gameplay, but it can’t yet construct fully playable games. Yet, as AI technology continues to advance, tools like Muse will transform game development and it is the players that will benefit from it.
