Google Genie 4 world model (research preview)

Google DeepMind announced Genie 4, a real-time world model that adds generated sound and dialogue; it is a limited research preview for now.

Google says Genie 4 builds on Gemini Omni, keeps scenes consistent for up to two minutes (double Genie 3), lets users change weather, camera and objects with text prompts, simulates physics and human movement more realistically, and is its first real-time world model with native audio, including dialogue. Gemini 3.8 Flash can act as a guiding agent for multi-step prompts. Output is watermarked with SynthID. Access is limited to trusted testers; Google says it will reach Google AI Ultra subscribers in Project Genie and early-access Gemini API and AI Studio partners in the coming weeks.

Status

Released

Date

Oct 28, 2026 (trusted testers; AI Ultra in the coming weeks)

Maker

Google

Type

AI model

Categories

AI

Sources

• Android Authority

Launch dates can change. Check the sources for the latest. Last reviewed Nov 10, 2026.