A generated landscape becomes a different creative tool when you can change direction and hear the world respond.
Google DeepMind introduced Genie 4 on October 27, adding native audio to its interactive world-generation system. The announcement describes environments people can navigate and influence while the model generates what happens next.
Access starts with trusted testers
Google has begun a research preview for trusted testers. Google AI Ultra subscribers are the first planned consumer group, through Project Genie. Early API and Google AI Studio partners are also part of the planned expansion.
That sequence matters for anyone planning a project around it. A waitlist and a named future access group do not establish that an ordinary Ultra subscription or API account can use Genie 4 today.
A scene can respond to the person inside it
A fixed video presents a chosen sequence of frames. An interactive world must respond when the user changes course. Google says Genie 4 combines that visual response with generated audio, making sound part of the same experience.
For a creator, the useful question is how much control survives after the opening scene. A beautiful first view leaves open whether a room remains recognizable after a detour or whether a sound follows the right event.
Evaluate a return journey
A practical evaluation could begin in one distinctive room. Leave a recognizable object there, walk away, then return and change the lighting or viewpoint. Record whether the object and room still agree with the earlier state.
Repeat the same journey while listening for sounds tied to movement and interaction. Keep visual persistence, responsiveness, and audio consistency as separate observations. This is a proposed evaluation method; we have not run Genie 4.
Google itself describes remaining limits in physical behavior, spatial consistency, object persistence, and text rendering. Those limits make the preview interesting for creative exploration while leaving reliable simulation for a separate assessment.
Follow the rollout
Use DeepMind’s announcement and waitlist to check access changes. For more developments in generated media, visit our AI Design coverage.
Which would you evaluate first in a generated world: returning to an earlier location or changing what happens inside it?
