Skip to main content
Tech

Genie 4 Is Out: Who Can Try Project Genie and What It Can Do

Google DeepMind's Genie 4 is now in Project Genie for eligible Google AI Ultra subscribers, with a waitlist for AI Pro. How to try it and what's new.

3 min read
Share
Illustration: glass orbs floating in dark space, the largest holding a miniature forest world at sunset

Google DeepMind has made Genie 4, its newest world model, available in Project Genie, the experimental app that turns a text or image prompt into an interactive world you can walk around in. For now it's for eligible Google AI Ultra subscribers, starting 18 November; Google AI Pro members can join a waitlist on the same page. Genie 4 lets you change a world while you're inside it, by typing a new prompt, and Project Genie adds an agent that builds worlds from a conversation, plus a gallery to share and remix them.

Who can try Genie 4

  • Google AI Ultra: eligible subscribers can use the new Project Genie experience from 18 November, in the browser at labs.google/projectgenie.
  • Google AI Pro: you can't use it yet, but you can join the waitlist on the same page.
  • Other plans: Google's post only mentions Ultra and Pro.

Google doesn't explain what makes an Ultra subscriber "eligible" or list the countries where it's available, so check the Project Genie page with your own account.

What Genie 4 adds

  • Change the world as you go: type a new prompt mid-session to shift the weather, drop in a character or redirect the action, and the environment updates in real time.
  • Better visuals, physics and audio: Google calls it its most realistic and interactive world model, with native audio.

Genie 3 is still there, and it got upgrades

Genie 3 stays in Project Genie's model picker for when you want longer runs:

  • Longer sessions: up to 3 minutes per generation, up from 60 seconds, and you can chain several worlds together.
  • 360-degree panoramas: seed a new world from a panorama image.
  • Live events: change the environment with natural-language prompts while you explore.
  • Agent: describe what you want in plain language and the Project Genie agent uses Google's models to build the world for you. You can also bring Gemini into a session to suggest where to go next or trigger events automatically.
  • Gallery: explore curated and community worlds, such as "Living Memory", which turns personal photos into explorable spaces, or remix one and publish your own for others to try.

Limits and safety

Google says Genie 3 and Genie 4 are still frontier research models, with open problems around camera latency, complex physical interactions, audio timing and keeping scenes consistent over time. All video and audio made in Project Genie carries a SynthID watermark, and inputs and outputs go through safety screening.

For more of Google's recent AI launches, see how Gemini can now do tasks for you for free and the Gemini 4 Hydrogen model behind them.

Source: Google, Google DeepMind: Genie 4