NewsPublished:
Google opens Genie 4 in Project Genie: AI worlds you can reshape while you play
Genie 4 reaches Google AI Ultra subscribers in Project Genie, adding live prompt edits, native audio, a world-building agent and a community gallery.

Google has started rolling out Genie 4 inside Project Genie, its browser-based prototype for generating worlds you can actually walk around in. Access goes first to eligible Google AI Ultra subscribers, while AI Pro members can join a waitlist and get in gradually.
What's new
The headline feature is that you are no longer stuck with the world you first described. While you explore, you can type a new prompt and change things on the spot: switch the weather, drop in a character, or push the scene in a different direction. Genie 4 also generates native audio in real time for the first time, from ambient sound to character dialogue.
The update brings a few more pieces:
- A Project Genie agent that builds a world from a conversation, setting its look and a matching background score.
- Gemini in the session, which watches what is on screen, suggests where to go next and can trigger in-world events.
- A community gallery where people can publish, explore and remix each other's worlds.
- Genie 3 upgrades: runs of up to 3 minutes per generation (up from 60 seconds), world chaining and 360-degree panorama seeding.
Google says everything made in Project Genie carries its SynthID watermark, and that both models are still frontier research systems.
My take
As someone who spends a lot of time in game worlds, the mid-session editing is the part that catches my eye. It feels closer to a sandbox toy than a game, and short sessions still limit what you can build. But being able to step into a scene, hear it, and reshape it on the fly is a big jump from the static clips we had a year ago. I'm curious whether creators use the gallery for fan-made settings, and how studios react.
Details are on Google's official announcement.
Source: blog.google