New to this? Read it in simple words
- Google has a new AI called Genie 4 that builds worlds you can explore.
- The worlds have sound, and characters in them can talk.
- You can change a scene, like the weather, by typing.
- For now, only trusted testers can use it.
- World model
- An AI that creates a virtual place and works out what happens next as you move through it.
- Research preview
- An early test version shared with a small group before a wider release.
- Gemini
- Google’s family of AI models.
Google showed Genie 3 a little over a year ago, and it reached paying users only through a limited web app, Project Genie. Genie 4 follows the same path, starting with a small group of testers.
What Genie 4 does
A world model creates a virtual place from a description, without pre-built scenes. It draws the world frame by frame in real time as the user explores it, Ars Technica explains.
With Genie 4, users move with the keyboard and type prompts to change the scene: the weather, the camera, or new people and objects. Every world also gets sound, including background noise, music and characters who talk.
Google says Genie 4 is the first real-time world model to make its own sound. It also says the model handles physics such as aerodynamics and collisions better, and that people in it move more naturally. These are Google’s claims.
From Google DeepMind’s announcement and reports of October 27, 2026.
Gemini in the loop
Genie 4 works with Google’s Gemini models. Gemini 3.8 Flash acts as a guiding agent: it breaks a goal into steps and checks each new frame, Google says. This helps a scene stay on track.
In a research test, Google’s Gemini 4 Argon model split robot tasks into steps that Genie 4 then simulated in real time. TechCrunch notes that this could let teams test robot scenes before trying them in the real world.
Testers have also used it to sketch ideas for films, a Google DeepMind engineer told reporters at a briefing.
Who can use it, and its limits
Only trusted testers can use Genie 4 now, and there is a waitlist. Google says it is coming soon to Google AI Ultra subscribers in Project Genie, and to early-access partners through the Gemini API and AI Studio.
TechCrunch reports that this will happen in the coming weeks; Ars Technica says there is no specific timeline. Google has not given a price, and AI Ultra plans start at $100 a month, Ars Technica notes.
In Project Genie, worlds could be explored for 60 seconds with Genie 3, and Genie 4 raises that to 2 minutes, Thurrott reports. Google says details and objects can still drift over time, and all video and audio carry its SynthID watermark.
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1
- 2Research · October 27, 2026Google’s Genie 4 world model adds audio, more realistic worlds, and real-time steering TechCrunch
- 3Research · October 27, 2026DeepMind unveils Genie 4 world model promising more realistic physics and humans Ars Technica
- 4Research · October 27, 2026Google Makes Genie 4 World Simulation Model Available for Select Partners Thurrott.com
- 5Research · October 27, 2026Google DeepMind launches Genie 4, CEO Sundar Pichai calls it ‘our most realistic and interactive world model’ The Times of India
The claims about realism, physics and a first for sound are Google’s own, and testers have not published results yet. Google’s own pages give no exact length for a consistent scene, so the 2-minute limit rests on one report.
Google’s Genie 4 builds worlds with sound you can steer. Silicon AI News, October 28, 2026. https://siliconainews.com/stories/google-genie-4-world-model
Genie 4 adds sound, talking characters and text steering to Google’s world model, a new ability held back by tester-only access and untested claims. How we rate the mood
The mood shows how the news frames the event for the tech and AI world: as an opportunity (6 and above) or as a loss (below 5). It is not our verdict on who is right.