Grafy News
Try Grafy free

Google DeepMind launches Genie 4 for interactive world creation

Google DeepMind rolled out Genie 4 to Google AI Ultra subscribers, adding real‑time text‑driven world editing.

Illustration for the news story: Google DeepMind launches Genie 4 for interactive world creationSource: Google DeepMind

What is Genie 4 and why does it matter?

Google DeepMind announced that Genie 4 is now available to eligible Google AI Ultra subscribers, extending the Project Genie platform with the most realistic and interactive world model to date. According to the Google DeepMind blog, Genie 4 lets users edit environments in real time simply by typing new prompts while navigating a world, enabling on‑the‑fly changes to weather, characters, and actions. The model also introduces native audio that reacts instantly to visual changes, turning a static scene into a fully immersive, multimodal experience.

Genie 4 is now available
Embedded audio from the source story.

How does Genie 4 differ from Genie 3?

Genie 4 focuses on high‑fidelity visuals, physics, and adaptive sound, while the earlier Genie 3 model remains available for longer sessions and panorama‑grounded worlds. The blog notes three key upgrades to Genie 3:

  • Extended sessions – Generations can now run up to three minutes, a jump from the previous 60‑second limit.
  • 360° panorama grounding – Users can seed new worlds from 360‑degree images, keeping visual consistency as the camera pans.
  • Real‑time in‑world events – Natural‑language prompts can dynamically alter the environment during exploration, linking multiple worlds into a continuous journey.

Genie 4 adds real‑time audio that adapts to scene changes, a capability not present in Genie 3.

How can creators use Genie 4 today?

Eligible Google AI Ultra subscribers can start building worlds immediately via the Project Genie website; Google AI Pro members may join a waitlist. The new Project Genie agent guides users through a conversational workflow: describe a vision, and the agent orchestrates multimodal models to generate visuals, background scores, and interactive elements. Creators can also bring Gemini models into the session for context‑aware suggestions and automatic event triggers.

The update includes a community gallery where users can publish, explore, and remix worlds created by others. Highlighted examples include:

  • Villa Giallo – a room‑by‑room mystery inside a gothic manor, built with Genie 4.
  • The Life of Ants – a photorealistic ant‑colony navigation experience.

These showcases demonstrate how real‑time editing and adaptive audio can support narrative storytelling, game prototyping, and educational simulations.

Safety and research considerations

DeepMind emphasizes responsible rollout. All video and audio generated by Genie 4 and Genie 3 carry a SynthID watermark, and every input and output passes through safety screenings. The blog acknowledges ongoing research challenges such as camera latency, precise physical interactions, real‑time audio fidelity, and long‑term scene permanence, directing readers to a detailed model research overview for technical depth.

Using Grafy to explore or build similar experiences

While Grafy does not yet host Genie 4 directly, creators can experiment with comparable workflows using existing apps:

  • Game Studio – Turn a brief description into a playable 2D/3D prototype, mirroring Genie’s “prompt‑to‑world” concept.
  • Creator – Generate high‑quality images or short video clips that can serve as assets for interactive scenes.
  • Graph – Track branching prompt histories and outputs, useful for managing the iterative edits that Genie 4 enables.
  • Community – Publish and remix your own interactive experiments, following the same social‑sharing model introduced by Project Genie.

For developers who want to dive deeper into the underlying research, Reader can ingest the Genie 4 model overview PDF and let you query the paper directly.


Frequently asked questions

Can Genie 4 edit a world while I am exploring it?

Yes. Genie 4 accepts natural‑language prompts during an active session, allowing you to change weather, add characters, or modify actions in real time, as described in the DeepMind announcement.

What new audio capabilities does Genie 4 provide?

Genie 4 pairs its visual output with native audio that adapts instantly to scene changes, delivering ambient soundscapes and character dialogue that evolve as the world is edited.

How long can a Genie 3 session run after the latest update?

The Genie 3 update extends generation length to up to three minutes per run, up from the previous sixty‑second limit, and also supports chaining multiple worlds together for longer experiences.


Source: Google DeepMind blog, “Genie 4 is now available”, 18 Nov 2026.


Source: Google DeepMind — Genie 4 is now available