← Wednesday, November 18 deck

05Product2 sources posting

Google: Genie 4, DeepMind's World Model, Is Now Available in Project Genie

Genie 4, our most realistic and interactive world model from @GoogleDeepMind, is now available in Project Genie 🌐 Here’s what’s new: Genie 4: Edit environments mid-session with text prompts, paired with real-time native audio and dialogue. Project Genie agent: Co-create worlds conversationally with custom soundtracks and live Gemini guidance. Community gallery: Discover, explore, and remix worlds from creators and the Project Genie community. Genie 3 upgrades: Longer sessions up to 3 minutes, world chaining, and 360-degree panorama grounding. Step inside what’s new ↓

Opening frame from Google's video introducing Genie 4 in Project Genie.

Context

Why it matters

A new official world model that can be edited mid session, with live audio and a helper agent, is a concrete step from world model research toward a product people can actually enter.

Google posted that Genie 4, which it called its most realistic and interactive world model from Google DeepMind, is now available in Project Genie. The same note said users can edit environments mid session with text prompts, with real time native audio and dialogue, and it listed a Project Genie agent, a community gallery, and Genie 3 upgrades including longer sessions, world chaining, and 360 degree panorama grounding. Google DeepMind posted a shorter first hand note at almost the same time, saying Genie 4 is now available in Project Genie and brings native audio, real time editing, and a new agent. Google's post is the lead because it is official, first hand, and the more complete product list, with the higher view count.

Deep dive

Google's official account said Genie 4, its most realistic and interactive world model from Google DeepMind, is now available in Project Genie. It listed what is new: edit environments mid session with text prompts, paired with real time native audio and dialogue; a Project Genie agent to co create worlds conversationally, with custom soundtracks and live Gemini guidance; a community gallery to discover, explore, and remix worlds; and Genie 3 upgrades of longer sessions up to 3 minutes, world chaining, and 360 degree panorama grounding. A video is attached. Those are first hand product claims from Google.

Google DeepMind posted a shorter first hand note one minute earlier, saying Genie 4 is now available in Project Genie and that it brings native audio, real time editing, and a new agent to help people create interactive virtual worlds. That lab post is official too, and it is folded here as additional reporting. Google's version is the lead because it names the same launch and adds the fuller feature list, and it had the higher view count when these posts were fetched.

The posts do not give pricing, access limits, or a quality comparison against Genie 3 beyond the listed upgrades. 'Most realistic and interactive' is Google's own wording. Mid session editing and native audio are the most important product claims, and they remain untested in this file.

Because both leads are official, the availability of Genie 4 in Project Genie, and the listed audio, editing, agent, gallery, and Genie 3 upgrades, are confirmed as Google's and DeepMind's announcement. How well the worlds actually hold up is not shown here.

What to expect next

A Project Genie page or a later DeepMind post should add who can use Genie 4, what the session limits are, and whether the agent and gallery are broadly open. Watch early creator tests of mid session edits, native audio, and the three minute Genie 3 sessions. If Google later publishes an evaluation against prior world models, that can be added. This should stay a Project Genie product story, not be merged into the Gemini 4 Neon release.