Google is rolling out a new Project Genie experience in your browser, and this time the worlds can change while you’re still walking around in them. Genie 4, Google’s latest world model, lets you type a new prompt mid-session to shift the weather, drop in a character, or redirect the action. It’s also the first version with real-time native audio, so ambient sound and character dialogue adapt as the scene unfolds.
You can try it now if you’re an eligible Google AI Ultra subscriber. That’s a narrower door than it sounds. Google AI Pro members can only join a waitlist on the same page, and the Google One Plan page says Project Genie is limited to the Ultra 20x plan, which costs $199.99 a month in the US. If you’re on the cheaper Ultra 5x tier, this one isn’t for you.
Genie 3 also gets a longer run
New world to explore, remix, and share
Genie 4 is the headline, but Genie 3 is still in the model picker when you want longer journeys or panorama-seeded spaces. Sessions now last up to 3 minutes, up from 60 seconds, and you can chain worlds together. You can also seed a world from a 360-degree panorama and change the environment with natural-language prompts while you explore.
Creation isn’t limited to a single text or image prompt anymore. A new Project Genie agent lets you describe your vision in conversation, then builds the world’s look and a custom background score to match the mood. Once you’re inside, Gemini can watch what’s happening on screen, suggest where to go next, or trigger in-world events as you move.
If you’d rather jump straight in, a new community gallery collects curated worlds and work from other users. You can explore, remix, and publish your own. The lineup includes Living Memory, which turns personal photos into explorable spaces, plus Dino Field Guide, Rock Paper Scissors, Villa Giallo, and The Life of Ants, made by filmmaker Anand Gandhi.
Creation isn’t limited to a single text or image prompt anymore.
Google says all video and audio generated in Project Genie carry SynthID watermarks, and inputs and outputs go through safety screenings. The models are still research tools, though, and Google is still working on camera latency, physics, audio precision, and long-term scene permanence. If you do get in, expect a few rough edges, not a finished game engine.