How to Access Genie 4: Google's New Interactive World Model

Google DeepMind announced Genie 4 on 27 October 2026. It is a world model that generates interactive, photorealistic environments you can navigate and reshape while they run. Unlike earlier video generators, Genie 4 produces visuals and audio together and responds to natural-language prompts in real time.

The model builds on the Gemini family. Gemini Omni anchors appearance and behaviour in real-world patterns. Other Gemini models supply the reasoning that keeps a simulation coherent as you move or change the scene. DeepMind has started a research preview for trusted testers. Wider access is planned for Google AI Ultra subscribers inside Project Genie and for selected early-access partners on the Gemini API and Google AI Studio.

What Genie 4 actually does

You start with a text prompt or reference image. The model then generates a navigable scene frame by frame. Standard keyboard controls handle movement and camera. Natural-language prompts let you alter the world on the fly: change the weather, move the viewpoint, add characters or objects, or adjust lighting.

Every world generates native audio. Ambient sounds, music, and interactive dialogue are produced with the visuals rather than added later. Official demos show characters that move more naturally and respond to environmental changes. DeepMind notes that visual details and object permanence can still drift over longer sessions or after many edits.

The company describes simulating the world as a step toward AGI. Potential uses include safe training environments for robots and agents, plus creative tools for filmmaking, journalism, gaming, and manufacturing. All output carries SynthID watermarking.

How access works right now

As of 28 October 2026, Genie 4 is available only to a limited group of trusted testers. Google is collecting feedback before broader release.

The stated next steps are:

1. Google AI Ultra subscribers will receive access inside Project Genie.
2. Early-access partners will gain access through the Gemini API and Google AI Studio.

No public date has been announced for either path. Project Genie currently still runs the previous Genie 3 model for Ultra subscribers. An Ultra subscription today does not guarantee immediate Genie 4 access.

The official waitlist form is linked from the DeepMind announcement page: https://forms.gle/WQnSNEM3omF7Uruu7

Steps to join the waitlist and prepare

1. Open the official DeepMind Genie 4 post or the waitlist form linked above.
2. Fill in the required details. The form asks for basic contact and interest information.
3. Submit and watch the email address you provide for updates.
4. If you already hold a Google AI Ultra subscription, keep it active. Project Genie is the planned consumer route.
5. For developers, monitor the Gemini API and AI Studio documentation. Early-access partners are invited separately; there is no public self-serve signup yet.

While you wait, you can review the official demo video above and the earlier Project Genie experience (still on Genie 3) if you have Ultra access. Prompting tips published for Project Genie remain useful: describe the environment in detail, specify character behaviour, and use clear language for camera or weather changes.

Practical limits and expectations

DeepMind has not published session-length numbers for Genie 4. Coverage of the demos notes that continuous interaction remains short compared with traditional game engines, and object permanence can degrade. Complex physics and long-horizon consistency are still active research areas.

Pricing for Genie 4 itself has not been announced. The only related consumer price is the existing Google AI Ultra plan, which starts at the higher tier of Google’s AI subscriptions. API pricing for Genie 4 has not been released.

If you are already working with Gemini models, the same reasoning capabilities that appear in Gemini 4 Argon and related releases are part of the loop that keeps Genie 4 coherent. See our earlier guide on adding Gemini 4 Argon to Android apps with Firebase AI Logic for related developer patterns: /blog/add-gemini-4-argon-android-firebase-ai-logic/.

Tips for when access arrives

• Start with a clear reference image or a detailed first-frame description. The model preserves key visual elements from the starting prompt.
• Use short, specific steering prompts rather than long paragraphs once the scene is running.
• Turn audio on. Native sound is a headline capability and helps judge coherence.
• Expect to restart sessions when details drift. Treat early access as a research preview, not a finished production tool.
• Check SynthID status on any output you plan to share.

Conclusion

Genie 4 advances real-time interactive simulation by combining photorealism, native audio, and Gemini-powered reasoning. Access is intentionally limited while feedback is gathered. Joining the official waitlist is the concrete step available to most people today. Developers should watch the Gemini API and AI Studio channels for early-partner invitations.

The model is still research software. Visual consistency over longer sessions and complex physical interactions remain open problems. When broader access begins, the combination of real-time steering and audio should open new workflows for both creative and simulation-based work.

Sources

• Google DeepMind: Genie 4: A new generation of world simulation (27 October 2026) — https://deepmind.google/blog/genie-4-a-new-generation-of-world-simulation/
• Google Blog: We’re announcing Genie 4 (27 October 2026) — https://blog.google/innovation-and-ai/models-and-research/google-deepmind/introducing-genie-4/
• Official demo video: Genie 4: A new generation of world simulation — https://www.youtube.com/watch?v=sFfPkUJd_Jk
• Waitlist form linked from the DeepMind announcement — https://forms.gle/WQnSNEM3omF7Uruu7