Genie 4: Google DeepMind's new world model adds sound and live editing

Google's Genie 4 is a "world model": an AI that generates an interactive scene from a text prompt and keeps simulating it as you steer. Google DeepMind released it on October 27, 2026 as a research preview for trusted testers, so most people can't try it yet.

According to TechCrunch's report on the launch, the new version adds sound, more realistic visuals and the ability to change a scene while it runs. This article relies on TechCrunch's account for the details; Google announced the model in its Genie 4 post and on the DeepMind site.

What is Genie 4?

Genie is Google DeepMind's family of world models. You describe a place, and the model generates it frame by frame in real time, reacting to what you do inside it. Genie 3 arrived as a research preview a little over a year ago; Genie 4 is the follow-up.

What's new compared with Genie 3?

Per TechCrunch, Genie 4:

• Generates audio that matches the world: music, ambient noise and interactive dialogue.
• Lets you edit mid-simulation by typing a prompt, for example changing the weather or camera angle, or adding people and objects.
• Holds a consistent scene for up to two minutes, double Genie 3's limit.
• Offers better physics and more humanlike characters.

Gemini is also more closely tied in. TechCrunch describes Gemini as a "director" that plans steps and checks frames while Genie draws the visuals. In one research experiment linked to Gemini Robotics, a Gemini 4 Argon model split a task into steps that Genie 4 simulated live. That points to a way of testing robot behavior in simulation before trying it in the real world. (We covered the model separately in Gemini 4 Argon is now available.)

Google research director Tejas Kulkarni told reporters the future is "hybrid world models," with Gemini writing code and text while Genie simulates reality, per TechCrunch.

Who can use Genie 4, and what does it cost?

Right now it is limited to trusted testers. In the coming weeks, Google says it will roll out to Google AI Ultra subscribers in Project Genie and to early-access partners of the Gemini API and Google AI Studio. TechCrunch's report gives no price and no exact date for wider access, so treat timing as open.

What are its limits?

Google's own team was candid. TechCrunch reports that details and objects can drift the longer a simulation runs, and that the team hasn't yet worked out how to guarantee accurate audio or legible in-world text. DeepMind engineer Diego Rivas also said it is still unclear what the finished product experience will be. A two-minute scene is a long way from a full game level.

Why does it matter?

World models are drawing money and rivals: TechCrunch notes that Runway released its second world model, GWM Worlds 2, a couple of months ago, and that the sector is starting to consolidate, with World Labs being acquired by AMD, Amazon reportedly in talks to buy Decart, and Rivian acquiring Third Dimension. Pitched uses include gaming, interactive learning and robotics. Trusted testers also see film prototyping, such as storyboarding, where Rivas said real-time generation helps keep creative "flow."

What does Genie 4 mean for you?

• Casual users: nothing to download today. If you pay for Google AI Ultra, watch Project Genie over the next few weeks.
• Developers: Gemini API and AI Studio access is planned for early-access partners first, not everyone.
• Creators: treat it as an idea sketchpad, not a production tool, until drift and audio issues improve.

For another real-time Google model, see our piece on Gemini 3.8 Live, which keeps talking while it works.

Sources

• Introducing Genie 4 (Google)
• Genie 4: A new generation of world simulation (Google DeepMind)
• Google's Genie 4 world model adds audio, more realistic worlds, and real-time steering (TechCrunch)