This video game doesn't exist until you play it. AI generates the world, characters and sound in real time.

Here's the full story, via Runway.

"GWM Worlds 2 is an autoregressive diffusion video and audio model generating 720p video at 24 fps and audio at 48,000 Hz." It's the successor to GWM Worlds, which Runway introduced in December, now extended with generated audio and richer subject and scene control.

"Both at once. A player crosses the desert as the survivor, walking, thrusting the spear, breaking into a run, while a director reshapes the scene around them: the sunset fades into a starry night and the campfire flares." Two people, two different kinds of control, over the same live-generated world, simultaneously.

"The world continues from each new input instead of following a fixed clip or script, so sessions have no preset length. This creates a foundation for interactive entertainment, virtual characters, robotics and embodied-agent simulation and generative design and interfaces."

The model runs on "WorldPrompt," Runway's format splitting a world into "what persists and what changes over time": a genesis prompt (scene, subjects, physical laws) plus a live, timestamped stream of text actions, movement, speech, sound, addressed to any subject or the scene itself. Runway describes three ways to use it: authoring the whole action sequence ahead of time (filmmaking), turn-based generation (interactive film), and true real-time play (games), the hardest case, since it needs near-zero-latency input.

Runway also tested agentic control, letting an AI agent steer a character and the environment directly. In one test, a simple robot agent was told to reach a red flag, then a blue flag, then both in sequence, "where it succeeded on its first try in both cases."

"GWM Worlds 2 is a research preview, and real-time generation still trades fidelity for speed. Difficult camera inputs such as very quick rotations can cause details, textures and geometry to degrade. Long-term memory is also imperfect, and the model doesn't support image references beyond the first frame or prefilled video and audio."

Runway is explicit this is an early point on a longer curve: "We expect the same curve that took offline video generation from short, rough clips to high-fidelity production footage to play out again here." It's not replacing game engines today, but it's a genuinely different proposal for what one could eventually be.

Source: Runway: https://runway.com/research/introducing-gwm-worlds-2

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter