AI NewsModels & agentsAnnouncement

Google opens Genie 4 to AI Ultra subscribers

Google AI Ultra subscribers can now use Genie 4, Google's latest world model, through Project Genie. Pro members join a waitlist.

AI News

Editorial2 min read

LinkedInX
Signal-field composition representing a Google world model.

Why it mattersA world model that takes new prompts mid-session is a new input for software that builds its own 3D scenes. Teams building game demos, robot training data or simulated workflows now have a vendor tool for it.

Google opened Genie 4 to Google AI Ultra subscribers on 18 November. Google AI Pro members can join a waitlist. Both reach Genie 4 through the Project Genie website at labs.google/projectgenie. Until now Genie 4 had been shown only in research posts. This is the first time the public can use it.

Genie 4 is Google's latest world model. Google calls it "our most realistic and interactive world model yet". A user starts a scene from a text prompt or an image. Google says native audio comes with it. The post names "ambient soundscapes to character dialogue" and says the audio adapts as the scene changes. A user can type a new prompt during a session. The model then adds characters, changes the weather or redirects the action, and reworks the scene in place.

Project Genie agent

Project Genie also opens a new agent. Google says the agent orchestrates its multimodal models. It turns a conversation into a first scene with a custom soundtrack. The agent then watches the user play and offers next steps. The post names "a city that comes to life, a dream that unfolds" as example prompts.

Genie 3 gets longer sessions

Google extended Genie 3 at the same time. A single generated scene can now run up to 3 minutes, up from 60 seconds. Genie 3 also adds 360-degree panorama grounding. A user can hand it a panorama and then explore it as a world. Chained worlds let the user walk between linked scenes. Real-time events can be triggered mid-session with a prompt.

Watermarking and safety

Every video and audio clip from Genie 4 or Genie 3 carries a SynthID watermark. Google says it screens inputs and outputs for safety. Users can share creations in a community gallery inside Project Genie. Other users can view and remix them.

Pricing, access and watermarks

A team can now build a simulated environment for an agent. It can also build a scene for a game prototype or a quick storyboard. The 3 minute session cap on Genie 3 is long enough for one short demo at a time. Chained worlds and panorama grounding give a way to stitch several scenes. The in-session prompt means a scene can be edited without a new generation.

Google states no API, no session count and no export beyond the watermarked video. Pricing stays tied to the existing Google AI Ultra and Pro plans. The SynthID watermark stays on every clip, so a generated scene cannot be passed off as filmed footage. Access is tied to the user's subscription, not a per-seat team plan. A studio or a research group that wants several seats pays one subscription per user.

Google says a Pro member who joins the waitlist "will be given access as capacity allows". That wording sets no date. A reader who wants Genie 4 today, and does not hold a Google AI Ultra plan, has nothing to install.

Source

Google: Project Genie Update: Genie 4 is now available

Google: Project Genie Update: Genie 4 is now available

SourceGoogle

This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.

Share
LinkedInX
Start a project