WHAM: World and Human Action Models
Unlocking new forms of creative expression and ushering in the future of interactive media
Our first World and Human Action Model published in Nature in 2025.
WHAM is the first instance of the World and Human Action Model. Trained on more than one billion images and controller actions from Ninja Theory’s Bleeding Edge, the equivalent of over seven years of continuous human gameplay, WHAM can generate complex, consistent gameplay sequences spanning several minutes from a one-second prompt. WHAM can be used in world-model mode (predicting how the game evolves), action-model mode (generating plausible controller actions), or both.
Try the model
Includes model weights, sample data, and the WHAM Demonstrator concept prototype.
Read the paper
Nature (2025) — World and Human Action Models towards gameplay ideation (opens in new tab)
Kanervisto, A., Bignell, D., Wen, L. Y. et al.
Read the announcement
Microsoft Research blog (2025) — Introducing Muse: our first generative AI model designed for gameplay ideation