Key takeaways
- Google began rolling out Genie 4 to eligible Google AI Ultra subscribers in Project Genie on Nov. 18, 2026, while AI Pro members can join a waitlist.
- Genie 4 generates explorable worlds with native audio and accepts new text prompts mid-session to change weather, add characters or redirect the action.
- Genie 3 stays available with sessions extended to 3 minutes from 60 seconds, plus panorama image grounding and prompt-triggered events.
In this story 4 sections
Google AI Ultra subscribers can now open Genie 4 inside Project Genie, Google's experimental app for building and exploring interactive worlds. Alphabet (NASDAQ: GOOGL) announced the rollout in a Google blog post on Wednesday, Nov. 18, saying eligible Ultra members get access starting that day. Subscribers on the cheaper Google AI Pro plan cannot use it yet, but they can sign up for a waitlist on the Project Genie page.
The release turns a research preview into something paying customers can try. Google DeepMind first described Genie 4 on Oct. 27, when access was limited to a small group of trusted testers. We covered that announcement in our first look at the Genie 4 world model.
What Genie 4 lets you do
Genie 4 is what Google calls a world model: you describe a scene in words, and the system generates a space you can move through with keyboard controls while it renders each moment on the fly. Google describes it as its most realistic and interactive world model so far.
Three features stand out in the company's description:
- Live edits. You can type a fresh prompt in the middle of a session to change the weather, drop in a character or send the action in a new direction, and the world updates while you play.
- Native audio. Genie 4 produces sound that fits the scene, from background ambience to dialogue spoken by characters. Google DeepMind called it the first of its real-time world models to generate audio itself.
- Gemini as a guide. A new Project Genie agent builds worlds through conversation, drawing on Google's other models for images and background music. It can also bring Gemini into a session to watch the screen, suggest what to try next or trigger events.
Genie 3 stays, with longer sessions
Genie 4 does not replace its predecessor. Genie 3 remains in the Project Genie model picker for users who want longer runs or worlds seeded from a panoramic photo, according to the post. Google also upgraded it:
- Each generation can now last up to 3 minutes, up from 60 seconds, a threefold increase, and worlds can be chained together.
- Users can start a world from a 360-degree panorama image.
- Plain-language prompts can set off events inside a running world.
Project Genie also gains a community gallery where people can publish worlds, browse what others made and remix them.
Limits Google lists
Google was direct about what still falls short. The blog post names camera latency, complex physical interactions, the precision of real-time audio and keeping a scene consistent over long stretches as open problems. The earlier DeepMind research post added that visual details can drift after several complex edits.
Neither post lists Genie 4's resolution, frame rate or a per-session time cap, and the rollout post does not name the countries where Ultra members can use it.
On safety, Google said every video and audio clip made in Project Genie carries a SynthID watermark, and all inputs and outputs pass through safety screening.
What it costs
Google did not list a separate price for Genie 4. Access comes with an eligible Google AI Ultra subscription, and plan details sit on the Google AI plans page. The rollout arrived the same day Google released its newest Gemini model; see our report on Gemini 4 Neon pricing and benchmarks, and more coverage in our AI section.
Share this story
Spotted an error? Tell us


