A world model you can walk through, retake with a sentence, and hear, from wind to spoken dialogue.
Google DeepMind introduces Genie 4 for real-time worlds you can steer while they run. Announced on 27 October 2026, it builds a scene from a text prompt, then lets you walk through it with keyboard controls. A new line can change the weather, move the camera, or add a person or object, and Gemini rewrites that request so it fits the scene already on screen. Between prompts, Gemini keeps the world moving. DeepMind frames it as a path to safe training grounds for robots and agents, and to tools for film, games, journalism, and engineering.
The leap from Genie 3 is sound. Genie 4 is DeepMind's first real-time world model that generates audio with the picture: wind, waves, music, and dialogue that reacts as the scene changes. Gemini Omni anchors how places and people look. Gemini 3.8 Flash watches the frames, splits a goal into steps, and checks that each step happened. The model simulates collisions, aerodynamics, and people who move more naturally and can be steered live. In one test, Gemini 4 Argon and Genie 4 simulated Apptronik's Apollo humanoid playing checkers, with Argon turning the task into captions Genie rendered as it ran.
Trusted testers in film, journalism, games, and engineering have it. Filmmakers use it as draft mode, stepping into a scene and changing camera, light, dialogue, and action. Photojournalists are testing whether one archival photograph can become a walkable version of that place. Game teams reach an explorable prototype before writing engine code. Every clip carries SynthID, Google's watermark for AI-generated video and audio. DeepMind says detail still drifts on longer runs, after many edits, or when you travel far. Readable text, exact audio, spatial accuracy, and long consistency are still open.
Genie 4 is a research preview for those testers. Everyone else can apply here. Google says the model is coming soon to Google AI Ultra subscribers inside Project Genie, and to early-access partners on the Gemini API and Google AI Studio. It has not published a Genie 4 price. The announcement is on Google DeepMind.
Original Sources
Disclaimer
This is an AI-summarized article from authoritative sources. If you want your content removed, please .
