Genie 4 opens to Google AI Ultra subscribers
Google has started opening Genie 4, its newest world model, to Google AI Ultra subscribers. A world model generates a place you can move around in from a text prompt. Genie 4 now runs inside Project Genie, Google’s experimental web app for building and exploring these worlds. The update also brings an agent that builds a world with you through conversation, a public gallery of worlds, and upgrades to the older Genie 3. Google announced the rollout on its official blog on November 18.
When Google DeepMind announced Genie 4 in late October, only trusted testers could use it as a research preview, and Ultra subscribers were told it would come “soon”. That wait is now over for Ultra. Google AI Pro members can join a waitlist on the same page.
What Genie 4 adds
The biggest change is that you can edit a world while you are inside it. As you move around, you type a new prompt and the scene changes in real time. Google’s examples are changing the weather, adding a character or sending the action in a new direction. Genie 4 also makes its own sound for the first time. The audio is generated live and follows the scene, from background noise to characters talking.
Genie 3 stays in the model picker for people who want longer runs. A single generation now lasts up to 3 minutes instead of 60 seconds, and you can link several worlds into one longer trip. Genie 3 can now also start a world from a 360-degree panorama photo, and you can change its world with a prompt while you explore.
The agent and the gallery
Before this update, you started a world from a single text or image prompt. With the new Project Genie agent, you describe what you want in plain language. The agent then uses Google’s other models to set the world’s look and write background music to match its mood. You can also bring Gemini into a session. It watches what is happening on screen, suggests where to go next and can set off events in the world by itself.
The gallery mixes worlds picked by Google with worlds made by users. Examples include “Living Memory”, which turns personal photos into places you can walk through, and “The Life of Ants”, made with filmmaker Anand Gandhi, which shows an ant colony from an ant’s point of view. You can remix any gallery world with the agent by changing its structure, soundtrack or rules, and then publish your own version for others to use.
Who can use it, and the limits
Google’s post says only “eligible” Ultra subscribers can use it. When Project Genie launched on January 29, it was open only to Ultra subscribers in the US aged 18 and over. At that time AdwaitX reported the Ultra plan cost $249.99 a month.
Google still calls both Genie 3 and Genie 4 research models. It says it is working on camera lag, complex physical interactions, the accuracy of live audio, and keeping a scene the same over long sessions. All video and audio made in Project Genie carries Google’s SynthID watermark, and everything going into and coming out of the models passes safety checks.
Up next
Safety camp and accelerationists unite against Anthropic
Accelerationists, AI safety researchers and Steven Pinker rarely agree. Ahead of Anthropic’s IPO, they signed one statement: it “is not a safe AI company.” Does Anthropic’s safety image play any part in why you use Claude?
Related stories
- Real-time world model Genie 4 generates its own audio
- Gemini Notebook reports generate slides on demand
- Avatars and a Customize tab head to Gemini desktop
- Sparkle watermark gets an off switch in Gemini
- Six countries get Gemini Tasks free, the EU doesn’t
- Gemini reads your Google Wallet spending, US users only
Comments
No comments yet. Be the first.
Join the conversation
Sign in to leave a comment. It appears after moderation.

