From the source
Google DeepMind is rolling out a new Project Genie experience to eligible Google AI Ultra subscribers, giving them first-time access to Genie 4 alongside upgrades to Genie 3, a creation agent and a community gallery.
- Early testers had asked for deeper control over what happens inside the generated worlds and easier ways to build and share them, the company says.
- Genie 4, described as its most realistic and interactive world model yet, lets users type a new prompt mid-session to edit environments in real time and pairs the visuals with real-time native audio that adapts as a scene unfolds.
- Genie 3 stays in the model picker for longer runs or panorama-seeded spaces, and now supports sessions of up to 3 minutes per generation, up from 60 seconds, world chaining, 360-degree panorama image grounding and real-time in-world events.
- The Project Genie agent turns a natural-language description into a world, orchestrating Google's multimodal models for initial visuals and custom background scores, while Gemini models can observe the screen and surface suggestion chips or trigger events.
- The gallery collects curated and community worlds, including collaborations with creative technologist Will aka BLVCKL!GHT and filmmaker Anand Gandhi, and users can publish their own creations for others to remix.
- All video and audio generated in Project Genie carries SynthID watermarking and all model inputs and outputs go through safety screenings, though the company says Genie 3 and Genie 4 remain frontier research models with open challenges in camera latency, physical interactions, audio precision and scene permanence.



