genie4

Genie 4 is Google DeepMind's natively multimodal world model, announced October 27, 2026. Instead of rendering a fixed clip, it builds a playable, physics-aware environment you can steer in real time with plain language, with ambient audio and characters that answer you.

The genie4 guide is an independent walkthrough of the announcement: what Genie 4 actually does, how prompting an interactive world differs from prompting a video, a side-by-side comparison with Genie 3, official demo clips, and the staged access picture (trusted testers first, then AI Ultra via Project Genie, plus Gemini API and AI Studio partners). It also includes a world prompt composer that helps you set the scene, motion, audio, framing, aspect ratio and clip length before generation, with example starting points such as a neon alley or an orbital station, and notes on film pre-visualisation, game prototyping and robotics training. Anything Google has not published, such as frame rate, output resolution and pricing, is listed as unverified rather than guessed.