Google Rolls Out Genie 4 in Project Genie for AI Ultra Subscribers

Google has begun rolling out Genie 4 inside a rebuilt Project Genie experience to eligible Google AI Ultra subscribers, letting them reshape interactive AI worlds with typed prompts mid-session while the model generates matching audio in real time. The update also adds a conversational creation agent, a community gallery for sharing and remixing worlds, and longer Genie 3 sessions, while Google AI Pro members can join a waitlist.

Nov 20, 2026 - 07:41
 0  1
Google Rolls Out Genie 4 in Project Genie for AI Ultra Subscribers
A person gestures toward a large screen showing a Google DeepMind Genie 4 fantasy world — a castle on cliffs above a river, a dragon flying overhead and a rider on horseback — with a prompt box reading "Describe a new scene..." and suggestion chips for changing the weather and adding underwater exploration.

Google Rolls Out Genie 4 in Project Genie for AI Ultra Subscribers

Google has started rolling out Genie 4, its most realistic interactive world model yet, to eligible Google AI Ultra subscribers inside a rebuilt Project Genie experience. The update, announced on November 18 by Google DeepMind and Google Labs, turns the experimental world builder into something closer to a living place: subscribers can now type a new prompt while they are still inside a world and watch the environment change around them in real time.

It is the first time Genie 4 has been available to consumers. DeepMind introduced the model on October 27 as a research preview for a limited group of trusted testers, saying wider access would follow. Three weeks later, that access has arrived — for the subscribers on Google's most expensive plan first.

Worlds that change while you walk through them

The headline change is mid-session prompting. Until now, a world in Project Genie was largely fixed once it was generated. With Genie 4, a subscriber exploring a scene can type a new instruction — shift the weather, drop in a character, redirect the action — and the world reshapes itself without restarting the session.

For the first time in the series, the visuals come with sound generated alongside them. Genie 4 produces real-time native audio that adapts as the scene unfolds, from ambient soundscapes to dialogue spoken by the characters it renders. That combination of moving images and matching audio is what DeepMind is positioning as the step from a generated video clip to a place you can inhabit.

Google is careful to frame both models as frontier research rather than finished products. The company says it is still working through camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence — the ways a generated world can still feel unsteady or forget what it showed a minute ago.

Genie 3 stays, with longer runs and panorama worlds

Genie 4 does not replace its predecessor. Genie 3 remains in Project Genie's model picker for subscribers who want longer runs or panorama-seeded spaces, and it gets upgrades of its own in this release.

Sessions on Genie 3 now run for up to three minutes per generation, up from 60 seconds, and multiple worlds can be linked together for longer journeys. Worlds can also be seeded from 360-degree panorama images, and Genie 3 gains its own real-time in-world events, letting explorers change the environment with natural-language prompts while they move through it.

A creation agent and a gallery for sharing worlds

The update also changes how worlds get built. A new Project Genie agent takes creation beyond a single text or image prompt: a subscriber describes a vision in natural language, and the agent orchestrates Google's suite of multimodal models to produce the opening visuals and a custom background score matched to the mood.

Gemini models can join the session as well, watching what is happening on screen, surfacing context-aware suggestion chips for where to go next, or triggering in-world events automatically as the journey unfolds.

Then there is the gallery. The new community space gathers curated worlds alongside creations from the community, and subscribers can publish their own worlds there for others to explore and remix. Google's examples include "Living Memory," which turns personal photos into explorable spaces; "Dino Field Guide," a prehistoric habitat built from 360-degree panoramas; and "Rock Paper Scissors," where themed objects face off on a desk with Gemini acting as referee.

The gallery also features collaborations with creators who had early access. Working with creative technologist Will, known as BLVCKL!GHT, Google built "Villa Giallo," a room-by-room mystery inside a gothic manor. With filmmaker Anand Gandhi, it created "The Life of Ants," which invites visitors to navigate a colony from an ant's perspective, drawing on Genie 4's photorealism.

Who can try it, and what it costs

Access goes to eligible Google AI Ultra subscribers, who can try the new Project Genie experience from November 18. Google AI Pro members cannot use it yet but can join a waitlist from the same page. Google has not published a Genie 4 session length for the consumer rollout, and the announcement gives no regional availability details, so the rollout appears to be reaching subscribers gradually.

Separately, Google also made Genie 4 available to developers through the Gemini API and Google AI Studio on November 19, with developer documentation pricing streamed output at about $0.10 per second and sessions of up to 120 seconds.

On safety, Google says all video and audio generated in Project Genie carries SynthID watermarking, and all model inputs and outputs pass through safety screening.

The Genie 4 rollout comes a week after Google introduced Gemini 4, its new flagship AI model family, as the company pushes its most capable models to paying subscribers first. For now, Genie 4's consumer debut is a test of whether people want to step inside AI-generated worlds at all — and whether a world that can be rewritten mid-sentence feels like a game, a film, or something new entirely.

Sources

Related Innopark IT article: Google Gemini 4 Hydrogen: Faster AI Model for Coding and Agentic Workflows

What's Your Reaction?

Like Like 0
Dislike Dislike 0
Love Love 0
Funny Funny 0
Angry Angry 0
Sad Sad 0
Wow Wow 0
Nahid Joy I’m Nahid Hasan Joy, a technology writer, web developer, and digital enthusiast with a strong interest in the ever-changing world of technology.