ai-toolsgeminitutorials

How to Prepare for Genie 4 in Project Genie: Prompts and Access

Practical prompts, keyboard controls, waitlist steps, and Ultra setup to get ready for Genie 4 real-time worlds in Project Genie.

October 28, 2026 Yum News

Google DeepMind announced Genie 4 on October 27, 2026. The model generates interactive, photorealistic worlds you can explore and reshape in real time, with native audio for ambient sound, music, and dialogue. Project Genie currently runs Genie 3 for eligible Google AI Ultra subscribers. Genie 4 arrives first for trusted testers, then for Ultra users inside the same Project Genie interface.

This guide shows how to prepare: join the waitlist, practice current Project Genie workflows, write prompts that transfer to the newer model, and master the keyboard controls. For a focused access overview, see our companion post on how to access the Genie 4 world model.

Current access and waitlist steps

Project Genie is an experimental Google Labs prototype. It is available to Google AI Ultra subscribers who are 18 or older in supported regions. Ultra plans start higher than Pro and include Project Genie alongside other advanced Gemini features. Street View grounding for real locations is limited to US spots in the current version.

Genie 4 is not yet in the consumer interface. DeepMind released it as a research preview to trusted testers and states it is coming soon to Ultra subscribers in Project Genie plus early-access partners on the Gemini API and Google AI Studio. No public timeline or separate pricing for Genie 4 has been published.

To join the waitlist:

Open the official DeepMind Genie 4 announcement or models page.
Click the “Join the Genie 4 waitlist” button. The form is at forms.gle/WQnSNEM3omF7Uruu7.
Sign in with your Google Account and complete the short fields about interest and intended use.
Submit. Approval times are not published; Ultra subscribers are prioritized for the Project Genie path.

Subscribing to Ultra now gives you the Genie 3 experience while you wait. Check the Project Genie interface or Gemini app for rollout notices once the model expands.

Earth from space highlighting global scale of simulation technology

How Project Genie works today with Genie 3

Open the Project Genie experience through Google Labs. High demand can place you in a queue; stay on the page. You have 30 seconds to define the world, then 60 seconds to explore the generated result.

Starting options include:

Text prompts for environment and character.
An uploaded image or Street View selection with style adjustments.
A random “roll the dice” starting point.

Choose first-person or third-person view. Nano Banana Pro generates a preview sketch so you can refine with add, change, or remove tools before entering. Generation finishes quickly and exploration starts automatically.

Keyboard controls you will use

Controls stay consistent between Genie 3 sessions and the Genie 4 demos shown by DeepMind. Standard keyboard input handles navigation while text prompts handle scene changes once steering is available.

Action	Key
Move forward / left / back / right	W / A / S / D
Jump or ascend	Spacebar
Turn or reorient the camera	Arrow keys

A progress bar shows remaining time. If the character moves backwards unexpectedly, press Spacebar to reset. Keep movements deliberate; short presses help when response latency appears. These controls appear in official help articles and in the Genie 4 announcement videos.

Prompt patterns that prepare you for Genie 4

Google’s March 2026 prompt tips for Project Genie remain the foundation. Structure every request around environment + character + movement style + viewpoint + sensory details. Keep language short and declarative.

Example for a current Genie 3 session:

Environment: A sunlit alpine meadow at golden hour with tall grass, scattered wildflowers, and distant snow-capped peaks. Soft wind moves the grass. Photorealistic lighting with warm highlights and soft shadows.

Character: A mountain biker in third person, riding a blue bike along a narrow dirt trail. Smooth forward motion, gentle turns, light dust kicked up behind the wheels.

For Genie 4 readiness, add continuity and audio notes even if the current model ignores some of them:

Specify ambient sound you expect (“soft wind through pines, distant water”).
Note possible mid-session changes (“allow weather shift to light rain on command”).
Ask for natural human movement and dialogue readiness (“character responds briefly when prompted”).

DeepMind demos for Genie 4 show real-time natural-language steering: change weather, move the camera, introduce a new person or object, or trigger dialogue without restarting. Gemini models help maintain coherence across frames. Practice concise follow-up instructions now so they feel natural later.

Official guidance recommends centering characters in reference images and leaving visible background. Use Gemini itself to refine long drafts into short, action-oriented sentences before pasting them into Project Genie.

Team collaborating around a laptop preparing digital creative work

Expected Genie 4 upgrades and how to adapt

Genie 4 is natively multimodal. It generates audio together with the visuals: ambient soundscapes, music, and interactive dialogue. Photorealism and physics improve, with better handling of aerodynamics, collisions, and natural human motion. Gemini Omni anchors appearance and behavior while other Gemini models plan multi-step events and check coherence.

Key practical differences once it reaches Project Genie:

Live steering via text while the world runs.
Native audio that matches the scene and responds to events.
Potentially longer continuous interaction than the current 60-second exploration window, though exact consumer limits are not published.
SynthID watermarks on outputs, consistent with other Google generative tools.

Watch the official DeepMind video for real-time examples of keyboard navigation plus prompt steering:

Setup checklist for Ultra subscribers

Confirm your Google AI Ultra plan and that you meet the 18+ age requirement in a supported region.
Join the Genie 4 waitlist using the official form linked from DeepMind.
Practice full sessions in current Project Genie: text prompts, image uploads, Street View where available, and both viewpoints.
Draft a library of short environment + character prompts that include movement style and one sensory detail.
Test keyboard fluency: WASD movement, Spacebar for vertical actions, arrows for camera.
Review the public gallery for stable prompt patterns others have used.
Note known issues from Google’s help pages (character control glitches, imperfect physics) so expectations stay realistic.

Privacy for Project Genie follows Google’s Private AI Compute approach for the prototype. Generated worlds carry SynthID. Treat the tool as an experimental creative and research environment rather than a production game engine.

Tips that improve results

Keep the first prompt focused. One strong environment and one clear character outperform long lists. Add only one major steering change at a time once live editing is available. Switch viewpoints mid-session to inspect details. If a reference image is used, ensure the character is centered and the background defines the space. Reuse successful prompts and regenerate; results vary. Check official help and the DeepMind blog for any session-length or feature updates, since the prototype continues to evolve.

Limitations persist. Complex long-horizon object permanence, perfect speech, and extended multi-minute consumer sessions remain research areas. Early Genie 4 demos show coherent scenes for short periods; independent measurements are not yet available.

Conclusion

Project Genie already lets Ultra subscribers create short interactive worlds from text, images, or Street View and explore them with simple keyboard controls. The same prompt structure—detailed environment, clear character, direct language, and viewpoint choice—prepares you for Genie 4. Joining the waitlist, practicing current sessions, and drafting audio-aware and steering-ready prompts puts you in position when the newer model reaches the interface.

Start with available Genie 3 sessions if you have Ultra access. Document what produces stable results. The combination of today’s controllable prototypes and Genie 4’s native audio plus live natural-language steering expands options for scene prototyping in film, games, education, robotics research, and creative exploration.