Google Genie 4 Arrives in Project Genie for AI Ultra Subscribers

Google Genie 4: archive photograph of a Google headquarters logo in Mountain View, California.

Google Genie 4 is now rolling out through Project Genie for eligible Google AI Ultra subscribers. Google announced the access update on November 18, 2026, adding its latest interactive world model, a conversational creation agent and a gallery for sharing and remixing projects.

This is an access update, not the original announcement of Genie 4. Google DeepMind introduced the model as a research preview on October 27. The newer rollout makes it available inside Project Genie to eligible subscribers, while Google AI Pro members can join a waitlist.

What Google Genie 4 adds in Project Genie

Project Genie is a browser-based experiment for creating and exploring generated interactive worlds. Google says Genie 4 lets a user change an environment while navigating it with a text prompt: for example, changing weather or introducing an object or character. The company also says the model generates adaptive native audio, including ambient sound and dialogue.

The update also includes a Project Genie agent intended to help build a world through conversation rather than a single prompt. A new community gallery lets users publish creations for other people to explore or remix. These are Google-described capabilities, not independently tested results.

Genie 3 remains available in the model picker. Google says its update extends a generated session to up to three minutes, adds linked worlds and permits a 360-degree panorama image as a starting point.

Who can use the Google Genie 4 rollout?

Access is rolling out to eligible Google AI Ultra subscribers through Project Genie. Google AI Pro subscribers do not receive the same immediate access; they can join the waitlist. That distinction matters for readers who already subscribe to a Google AI plan: availability depends on the plan and eligibility, rather than simply having a Google account.

The practical appeal is easier experimentation with interactive scenes, but it should not be confused with a conventional game engine or a production-ready simulation platform. Google describes the work as frontier research and says it is still addressing camera latency, detailed physical interactions, real-time audio precision and long-term scene permanence.

Useful, but still a research product

DeepMind says Genie 4 can lose visual consistency or object permanence during longer sessions, after several complex edits or while moving through a world. It also lists legible text, spatial accuracy and complex interactions among its open problems. Google says generated video and audio include SynthID watermarking and that inputs and outputs receive safety screening.

For now, the sensible reading is that Project Genie offers eligible subscribers a new creative and experimental tool, not proof that every generated world will behave predictably. Readers following practical AI releases can find broader updates in DITD’s News archive.

Google’s rollout details are available in its Project Genie update; DeepMind’s Genie 4 research overview explains the model’s stated capabilities and limitations.

Archive photograph of a Google headquarters logo in Mountain View, California, photographed by Anthony Quintano in 2014. Photo: Anthony Quintano from Mount Laurel, United States / CC BY 2.0. Resized/display crop; original image license applies.

Scottshak

I’m an automation architect and the dude behind these guides. Away from the keyboard, I write poetry, books and screenplays, and love making films.

You may also like...

Leave a Reply