Google DeepMind announced Genie 4 on October 27, 2026, introducing an interactive world model capable of generating real-time simulations with native audio and physics.
Multimodal Simulation and Controls
Genie 4 is natively multimodal, generating ambient sound, music, and interactive character dialogue simultaneously with visual environments. The system allows users to steer simulations in real time using standard keyboard navigation, camera controls, and natural language prompts.
The model builds on Google's Gemini family of models. DeepMind said Gemini Omni anchors simulations in real-world appearance and behavior, while other Gemini models provide reasoning to maintain continuity across scene changes.
Reasoning and Robotics Workflows
To manage multi-step actions in real time, Gemini 3.8 Flash operates as a guiding agent that observes generated scenes, breaks down user goals into steps, and tracks frame progress as the environment shifts.
DeepMind also reported testing Genie 4 with Gemini 4 Argon to simulate robotic tasks, including an experiment breaking down steps for Apptronik's Apollo humanoid playing checkers. Early testers across filmmaking, game design, journalism, and enterprise simulation have evaluated the tool for concept design and rapid environment prototyping.
Safety, Availability, and Limitations
Genie 4 incorporates internal safety evaluations and embeds Google's SynthID digital watermark across all generated video and audio to identify synthetic media.
DeepMind noted current limitations, explaining that visual details and object permanence can drift during extended sessions, navigation, or frequent prompt edits. Genie 4 is currently available as a research preview for trusted testers and is planned for Google AI Ultra subscribers in Project Genie, alongside Gemini API and Google AI Studio early access partners.
What to do
- Users interested in testing Genie 4 can join the Project Genie waitlist to access future releases for Google AI Ultra subscribers.
- Developers can seek future access through early access partner programs for the Gemini API and Google AI Studio.
Key facts and where they come from
- Genie 4 generates ambient sounds, music, and interactive dialogue directly alongside visual scenes.
Genie 4 is natively multimodal, enabling audio-like music, ambient noises, and interactive dialogue to be simultaneously generated with the world.
- Gemini 3.8 Flash acts as an agent in the loop to guide multi-step scene progression during live simulation.
To achieve this in real-time, Gemini 3.8 Flash serves as a guiding agent in the loop: as Genie 4 generates a world in real-time, Gemini continuously observes the scene
- All media generated by Genie 4 includes SynthID watermarking across audio and video outputs.
all video and audio generated by Genie 4 will include SynthID, our imperceptible digital watermarking technology that helps identify AI-generated content across visual and audio modalities.
- DeepMind has released Genie 4 as a research preview before rolling it out to Google AI Ultra subscribers.
We have started rolling Genie 4 out as a research preview for trusted testers — gathering feedback before we release it to consumers and developers, starting with Google AI Ultra subscribers
