Article Complete
AIM
Microsoft Used AI to Recreate Quake II
Microsoft Used AI to Recreate Quake II
Nostalgia Overload

Microsoft Used AI to Recreate Quake II

Microsoft used its AI model for video games to build a gameplay demo of Quake II.

Published Updated

Key takeaways

  • Microsoft unveiled an AI-generated gameplay demo of Quake II using its Muse model.
  • The demo allows players to experience Quake II in real-time within a browser.
  • Microsoft’s WHAMM model trained on a week of curated gameplay, reducing training time significantly.
  • The project showcases advancements in AI for video game development and real-time world modeling.
  • Quake II, a classic FPS game from 1997, is now reimagined through cutting-edge technology.
5 min

On April 4, Microsoft unveiled an interactive gameplay demo in Copilot Labs. It is an AI rendition of Quake II powered by Muse, its generative AI model for video game visuals.

Quake II is a first-person shooter (FPS) game published in 1997 by Activision Blizzard, which is now owned by Microsoft.

To make this possible, Microsoft introduced a real-time playable extension of their previous model, i.e, World and Human Action MaskGIT Model (WHAMM). 

The process involved training WHAMM on approximately one week of curated gameplay data from Quake II, a significant reduction from the seven years of data used for its older model to generate similar results. This focused data collection, aided by professional game testers, allowed for training on a single level with intentional gameplay. 

https://youtu.be/4OVcVG52hGA

The blog post explained the experience in brief, saying, “Much to our initial delight, we were able to play inside the world that the model was simulating. We could wander around, move the camera, jump, crouch, shoot, and even blow-up barrels similar to the original game.” It added, “Additionally, since it features in our data, we can also discover some of the secrets hidden in this level of Quake II.”

Microsoft stated some of its limitations, including that enemy interactions are often fuzzy and combat can be inaccurate. The short context length (0.9 seconds) leads to the model "forgetting" objects out of view, causing inconsistencies. Counting, particularly health values, is unreliable. The experience is limited to a single part of one Quake II level.

Finally, making WHAMM widely available introduces noticeable latency. Microsoft emphasises that this is a research exploration to understand the potential of AI in creating new interactive media.

Membership

Stay ahead with AI insights.

Unlimited access to premium articles, exclusive analysis, and the stories shaping India's AI economy.

Subscribe now

AIM Americas

1190 Miraloma Way Suite P, Sunnyvale, CA 94085

AIM India

1st Floor, Sakti Statesman, Bellandur, Bengaluru – 560103

© 2026 AIM. All rights reserved.