Genie 4 lands in Project Genie: rewrite weather and characters mid-walk, with real-time native audio for the first time
Synopsis
Google opened a new Project Genie experience to eligible AI Ultra subscribers, offering Genie 4 for the first time alongside Genie 3 upgrades. Early testers could generate and explore interactive worlds but wanted deeper control over what happens inside them; Genie 4 lets users type a new prompt mid-session to edit the environment in real time and pairs visuals with real-time native audio for the first time. Genie 3 sessions extend from 60 seconds to up to 3 minutes, with world chaining and 360-degree panorama seeding.
Interpretation
Interactive worlds in Project Genie can now be rewritten during exploration, with a typed prompt shifting the weather, dropping in a character or redirecting the action. Early testers asked for deeper control over what happens inside those worlds and easier ways to build and share them. The capability arrives with first-time access to Genie 4 in Project Genie, rolling out to eligible Google AI Ultra subscribers.
Genie 4 pairs its visuals with real-time native audio that adapts as the scene unfolds, from ambient soundscapes to character dialogue. Genie 3 provided world generation and exploration; Genie 4 is described as the most realistic and interactive world model yet. Genie 3 remains in the model picker for longer runs or panorama-seeded spaces.
Genie 3 sessions extend from 60 seconds to up to 3 minutes per generation, and multiple worlds can be linked into longer journeys. The previous per-generation limit was 60 seconds. The same Genie 3 update adds 360-degree panorama images as seeds for new worlds and natural-language prompts that change the environment during exploration.
A new Project Genie agent turns creation into a conversation: describe a vision in natural language and the agent orchestrates Google's multimodal models to craft initial visuals and a custom background score. Creation is no longer limited to a single text or image prompt. Gemini models can also join the session, observe the screen, surface context-aware suggestion chips or trigger in-world events automatically.
Perspective
The demonstrated scope is the new Project Genie experience for eligible Google AI Ultra subscribers, covering Genie 4, Genie 3 updates, the Project Genie agent and a community gallery; Google AI Pro members can join a waitlist on the same page.
Genie 3 and Genie 4 remain world models in the frontier research space, and Google says it is actively tackling camera latency, intricate physical interactions, real-time audio precision and long-term scene permanence. All generated video and audio carry SynthID watermarking, and model inputs and outputs go through safety screenings.
