█

AI/TLDR

Google DeepMind · 2026-11-18 · major

Genie 4 — DeepMind's world model now live in Project Genie

In short

Genie 4 is Google DeepMind's real-time world model with native audio. It is rolling out in Project Genie to eligible Google AI Ultra subscribers, and Google AI Pro users can join a waitlist.

Google DeepMind · 2026-11-18 · 4 sources checked

Project Genie thumbnail for the Genie 4 world model launch

Genie 4 lets you step into a generated world, with sound, and change it by typing new prompts while you explore.

Quick facts

MakerGoogle DeepMind and Google Labs
ProductProject Genie
AvailabilityEligible Google AI Ultra subscribers
WaitlistGoogle AI Pro
AudioNative real-time audio
Genie 3 session length3 minutes (was 60 seconds)
WatermarkSynthID on video and audio

What is it?

Genie 4 is Google's most realistic and interactive world model so far. It is now available inside a new Project Genie experience, where users can type prompts mid-session to change the weather, add characters or redirect the action.

How does it work?

DeepMind's research post says Gemini adapts each prompt to fit the scene. A Project Genie agent builds worlds from a chat description and adds custom soundtracks, and Gemini can watch a session to suggest where to go or trigger events. Genie 4 is the first model in the line to pair visuals with real-time native audio.

Why does it matter?

People can now try a world model in the product, not only watch research demos. Genie 3 stays in the model picker for longer runs and for worlds seeded from 360-degree panoramas, and its sessions now last 3 minutes instead of 60 seconds.

Who is it for?

Google AI Ultra subscribers, creators and world-model researchers

Frequently asked questions

Who can use Genie 4 in Project Genie today?
Genie 4 in Project Genie is rolling out to eligible Google AI Ultra subscribers. Google AI Pro members can join a waitlist on the announcement page. The post gives no price for Project Genie and does not list a rollout date for other plans.
How is Genie 4 different from Genie 3?
Google calls Genie 4 the highest-fidelity option, with the best visuals, physics and audio, and the first in the line with real-time native audio. Genie 3 remains in the model picker for longer runs and panorama-seeded spaces, and it now generates up to 3 minutes per session.
Can Genie 4 worlds be shared?
Project Genie includes a community gallery where users can publish, explore and remix worlds. Google lists examples named Living Memory, Dino Field Guide and Rock Paper Scissors. Genie 4 output carries SynthID watermarks on video and audio.
What are the known limits of Genie 4?
Google names camera latency, complex physical interactions, real-time audio precision and long-term scene permanence as open challenges for Genie 4. DeepMind's research post adds that visual details can drift over long sessions or after several prompt edits.

Try it

https://labs.google/projectgenie

Sources · 4 outlets

Tags

  • genie-4
  • world-model
  • google-deepmind
  • project-genie
  • google-labs
  • audio
  • simulation

← All releases · Learn AI