Google has begun rolling out a new Project Genie experience, built around Genie 4, to eligible Google AI Ultra subscribers. Genie 4 is the company’s most realistic and interactive world model yet, according to Google, and the update arrived on November 18, 2026. Google AI Pro members can join a waitlist.

Project Genie turns a text or image prompt into an interactive world that you explore from your browser. The new version lets you change that world while you are inside it, and adds a creative agent and a community gallery.

What can Genie 4 do that Genie 3 could not?

Genie 4 lets you type a new prompt in the middle of a session and watch the world change.

  • Shift the weather, drop in a character or redirect the action in real time
  • Native audio that adapts as the scene unfolds, from ambient soundscapes to character dialogue
  • Gemini models that watch what is on screen, offer suggestion chips for where to go next, and trigger in-world events

Google says Genie 4 pairs its highest-fidelity visuals and physics with audio for the first time.

Genie 4 sample scene of a woman with pink-streaked hair looking out of a car window at night in a city
Photo: Google DeepMind

What changed for Genie 3?

Genie 3 stays in the model picker for people who want longer runs or panorama-seeded spaces. Google’s updates to it include:

  • Sessions of up to 3 minutes per generation, up from 60 seconds
  • World chaining, which links several worlds into a longer journey
  • 360-degree panorama image grounding, to seed a world from a panorama
  • Real-time in-world events triggered by natural-language prompts
Genie 4 sample scene of a pink ribbon of fabric winding through a grassy hillside
Photo: Google DeepMind

The Project Genie agent builds a world from a conversation rather than a single prompt. Google says it orchestrates its multimodal models to create the starting visuals and a custom background score that fits the mood.

The new gallery mixes curated worlds with community creations. Google names “Living Memory,” which turns personal photos into explorable spaces, “Dino Field Guide,” and “Rock Paper Scissors,” where Gemini acts as a real-time referee. Two collaborations are also featured: “Villa Giallo,” a room-by-room mystery in a gothic manor, and “The Life of Ants,” made with filmmaker Anand Gandhi. You can remix gallery worlds and publish your own.

Genie 4 sample scene of a wooden rowboat seen from above on a pond with water lilies
Photo: Google DeepMind

How is it kept safe?

Google says all video and audio made in Project Genie carries SynthID watermarking, and that model inputs and outputs go through safety screenings. It also lists open research problems: camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence.

Genie 4 sample scene of a marble bust on a stone pedestal in a garden with slices of watermelon
Photo: Google DeepMind

Prices

Google’s announcement does not list a price. Access comes through a Google AI Ultra subscription, and Pro members are waitlisted. TokenPost puts AI Ultra at $249.99 per month in the United States, but an October report from Digital Trends gave a different monthly figure, so confirm the current price on Google’s plans page for your country before you subscribe.

Genie 4 sample scene of a station worker in uniform on a platform beside a red train
Photo: Google DeepMind

Who it suits

Creative experimenters on AI Ultra. If you already pay for the tier, the prompt-to-world flow, the agent and the gallery give you something to try today, and remixing gallery worlds is a gentle way in.

Game developers and filmmakers. Digital Trends reported that Google points to game prototyping and filmmaking as intended uses. Treat Genie 4 as a sketching tool, since Google itself calls these world models frontier research.

AI Pro members. You can join the waitlist on the Project Genie page, but Google gives no date for opening it up.

Readers who just want a look. The gallery is the quickest route to see what others have built, though the sources do not say whether it can be browsed without a qualifying subscription.

Genie 4 sample scene of a crowned crane walking across snow
Photo: Google DeepMind

What is still unknown?

Google does not define who counts as eligible, and TokenPost notes that the sources do not cover availability outside the United States. The post does not give a session length for Genie 4 itself; Digital Trends and Implicator, covering the October research preview, said experiences lasted about two minutes. Google also says simulations can drift in detail and object permanence over longer sessions.

Next step: open the Project Genie page from Google’s announcement to check your plan’s access, and watch the official Genie 4 video from Google DeepMind to see what to expect first.