WHAM: World and Human Action Models
Unlocking new forms of creative expression and ushering in the future of interactive media
Our real-time World and Human Action Model, now available to play as a technical demo in Copilot Labs.
WHAM-RT (World and Human Action Model — Real Time) is our real-time WHAM. By moving from autoregressive token-by-token generation to a MaskGIT-based approach, WHAM-RT generates visuals at 10+ frames per second, fast enough to play inside the model in real time. WHAM-RT also transferred the WHAM recipe to a new game, Quake II, using only one week of carefully curated gameplay data (compared to the seven years used for Muse), and doubled the output resolution to 640×360. WHAM-RT powers an interactive technical demo available to play in Copilot Labs.
Try it now
Read the article
Microsoft Research blog (2025) — WHAMM! Real-time world modelling of interactive environments.