On this page
Is Genie 4 now available?
Yes, but only for some people. Google rolled out Genie 4, its real-time world model from Google DeepMind, on 18 November 2026 to eligible Google AI Ultra subscribers through the browser-based Project Genie. AI Pro members can join a waitlist, and Google has not said when they will get access. Genie 4 was first shown to trusted testers in October.
What did Google announce on 18 November 2026?
In its Project Genie update post, Google said it is "rolling out a new Project Genie experience to eligible Google AI Ultra subscribers". It describes Genie 4 as the most realistic and interactive world model it has released. Eligible users can try it at labs.google/projectgenie, and AI Pro members can join a waitlist on the same page.
Genie 4 generates a world as you move through it, rather than rendering a finished video from a prompt. Decrypt and The Next Web both report that the announcement came on Wednesday 18 November, which is why it was still circulating on Hacker News on 19 November.
What can Genie 4 do?
Google lists these headline features for Genie 4:
- Mid-session edits. You can type a new prompt while exploring to change the weather, add a character or redirect the action.
- Native audio. Google says it is the first Genie model to pair visuals with real-time audio that adapts to the scene, including ambient sound and character dialogue.
- A Project Genie agent. It builds worlds from a conversation, creates matching background music, and lets Gemini suggest where to go next or trigger in-world events.
- A community gallery. Users can remix curated and community worlds, then publish their own.
Google also names two worlds built with early-access creators: "Villa Giallo", a gothic-manor mystery, and "The Life of Ants", made with filmmaker Anand Gandhi.
What happens to Genie 3?
Genie 3 stays in the model picker. Google says its sessions now last up to three minutes per generation, up from 60 seconds, and that worlds can be chained into longer journeys. New worlds can also be seeded from 360-degree panorama images. Decrypt frames the choice as Genie 3 for longer explorations and Genie 4 for stronger visuals and audio. Neither Google nor Decrypt gives a session limit for Genie 4.
Who can use it, and what does it cost?
| Question | What the sources say |
|---|---|
| Who has access now? | Eligible Google AI Ultra subscribers |
| Who is on a waitlist? | Google AI Pro members |
| Price | Decrypt reports AI Ultra costs $199.99 a month in the United States; Google's post does not state a price |
| Regions | Not specified in Google's post or the coverage we read |
| Usage caps for Genie 4 | Not stated |
"Eligible" is not defined in the coverage we read, so UK readers should check the Project Genie page before assuming they can use it. We could not confirm UK availability.
What did Google say in October?
Genie 4 was first announced as a research preview. Fone Arena's report, dated 28 October 2026, says Google was working with trusted testers in creative industries, journalism, education, gaming and engineering, and that enterprise partners were evaluating it for industrial simulation. It also says Gemini 3.8 Flash guides complex event simulation by breaking goals into steps. Google said at that stage that the model would come to AI Ultra subscribers through Project Genie and to early-access partners through the Gemini API and Google AI Studio. The 18 November update covers the first of those routes only. The sources we read do not mention API access as part of this release.
What are the limits?
Google calls both Genie models research systems. Its post says it is still working on camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence. Decrypt adds that objects can drift or change between views, complex physics remain difficult, and generated speech or text may be inaccurate. It also notes that improvements to movement and physical responses come from Google's own demos rather than independent testing.
Google says all video and audio generated in Project Genie carries SynthID watermarking and that inputs and outputs are safety screened.
Analysis: why this release matters
This section is our own reading, not reporting.
World models are usually judged by whether they hold a scene together while a person interacts with it. Genie 4's selling points, editing a world mid-session and hearing it respond, are interactive features, so the real test is how long a session stays coherent. Google has not published a Genie 4 session length, and its own limitations list names scene permanence as unfinished work. Until independent users report on longer sessions, treat the demos as vendor claims.
The release also follows a familiar Google pattern: a research preview to testers, then a paid tier, then a waitlist for the cheaper tier. Our coverage of Gemini 4 Carbon opening to paying users and of Gemini 4 Hydrogen in Google Search shows Google's other recent launches in context.
The Project Genie agent is the part worth watching. It lets Gemini observe the screen and trigger events, which moves Genie from a model you prompt towards a system that acts. If that idea is new to you, see our explainer on what agentic AI is. The Next Web also places the launch against rivals: it notes AMD agreed in September to buy World Labs for $8.2bn, and that Odyssey says its Odyssey-3 model can control robots and cars. We have not verified those two points independently.
What should you do next?
- If you subscribe to AI Ultra, open Project Genie and check whether Genie 4 appears in your model picker.
- If you are on AI Pro, join the waitlist; Google has given no date.
- If you are a developer, note that the update post does not announce Gemini API access.
Sources
- Google: Project Genie update featuring Genie 4
- Decrypt: Google's Genie 4 Is Here, Letting Ultra Subscribers Build AI Worlds in Real Time
- The Next Web: Google opens its Genie 4 world model to AI Ultra subscribers
- Fone Arena: Google DeepMind rolls out Genie 4 with real-time simulation and native audio generation
FAQ
Frequently asked questions
Google's update post says Genie 4 is rolling out to eligible Google AI Ultra subscribers, which is a paid plan. Decrypt reports AI Ultra costs $199.99 a month in the US. We found no free access route.
Google announced the Project Genie rollout on 18 November 2026. Genie 4 was first revealed as a research preview for trusted testers in late October 2026.
Google's post does not list regions, and the coverage we read does not define who is eligible. Check the Project Genie page. We could not confirm UK availability.
Per Google and Decrypt, Genie 4 offers stronger realism, mid-session prompt edits and native audio including dialogue. Genie 3 remains available for longer explorations, now up to three minutes per generation.
No. Decrypt, summarising Google, says Genie is not a traditional game engine with fixed rules. It predicts how a world should look and sound as it changes.
Spotted an error? Tell us. Read how we report and correct stories in our editorial policy.
Written by
Mainstream Tech
Newsroom
News and analysis on technology, AI, business, politics and international affairs, from our newsroom in London.
More from Mainstream




