Skip to main content
Google DeepMind's Genie 4 update: chainable AI worlds, remix gallery. A futuristic interface displays interconnected virtual

Editorial illustration for Google Deepmind's Genie 4 update adds chainable worlds, remix gallery

Google DeepMind's Genie 4 Builds Interactive 3D Worlds

Google Deepmind's Genie 4 update adds chainable worlds, remix gallery

• 4 min read

Google DeepMind is rolling out Genie 4, the newest version of its world model, moving it from a research preview into the hands of paying users. The model builds interactive 3D worlds that respond to text commands as they run, letting someone swap out weather, characters, or the direction of a storyline mid-session. It also generates sound in real time that matches whatever is happening on screen, rather than layering in pre-made audio.

Genie 4 lands as part of a broader update to Project Genie, which includes fresh work on the older Genie 3 model: sessions now stretch to three minutes, 360-degree panoramas are supported, and worlds can be chained together into longer sequences. DeepMind is also adding an AI agent that builds worlds straight from spoken descriptions, plus a gallery where users can post their creations for others to remix.

Access is split by subscription tier. Google AI Ultra subscribers get Genie 4 now, while Google AI Pro members can sign up for a waitlist. Every output carries a SynthID watermark and runs through safety checks before release, a detail DeepMind is framing as central to how it's opening up the tool.

Google Deepmind is releasing Genie 4, its most advanced world model for generating interactive 3D worlds. The new version, previously available only as a research preview, is part of a major update to Project Genie and is now open to Google AI Ultra subscribers. Genie 4 lets users reshape worlds through text input in real time, changing weather, characters, or storylines.

Why this matters

Genie 4 is the first version of this world model that actual paying users, not just research partners, get to put their hands on. That's the real shift here: Google AI Ultra subscribers can now generate and reshape 3D worlds with text prompts, chain them together, and remix what other people built in a public gallery. For developers and founders watching the world-model race, the gallery feature is worth tracking closely.

It turns Genie from a closed demo into something with a feedback loop, where prompts and edits get reused and built on in public. That's how you find out fast whether a generative world tool has staying power or just novelty. We're skeptical of "real-time" claims until independent users stress-test latency and consistency across longer sessions, and a gating behind Ultra subscriptions means broad judgment will lag the headline.

Still, putting chainable, editable worlds in front of a real user base, rather than a research sandbox, is the kind of move that tells you Google thinks this category is close to commercial, not just academic.

Common Questions Answered

What are the key features that distinguish Genie 4 from previous versions of Google DeepMind's world model?

Genie 4 introduces chainable worlds that can be connected together, a remix gallery for sharing and building upon other users' creations, and real-time sound generation that dynamically matches on-screen events rather than using pre-made audio. The model also allows users to reshape interactive 3D worlds through text commands, enabling them to swap out weather, characters, or storyline directions mid-session without interruption.

Who can now access Genie 4 and what is the significance of this availability shift?

Genie 4 is now available to Google AI Ultra subscribers, marking the first time this world model has moved beyond research partners and closed demos into the hands of paying users. This represents a major shift from research preview status to a commercially available product, making advanced world-generation capabilities accessible to developers and founders interested in exploring the technology.

How does Genie 4's real-time audio generation work compared to traditional world-building tools?

Genie 4 generates sound in real time that matches whatever is happening on screen, rather than layering in pre-made audio tracks as traditional tools do. This dynamic audio generation creates a more immersive and responsive experience where the soundtrack adapts to the interactive 3D world's changes as they occur.

What role does the remix gallery play in Genie 4's broader ecosystem?

The remix gallery is a public feature that allows users to share their generated worlds and enables other users to remix and build upon those creations. This gallery transforms Genie from a closed demonstration tool into a collaborative platform, making it particularly valuable for developers and founders tracking the world-model race.

LIVE00:45Auditors Struggle to Catch Sabotage in Full AI Safety Research Transcripts