genie-4

The tweet claims Google has released Genie 4, capable of generating photorealistic AI worlds explorable in real time, with support for free movement, talking to characters, and changing scenes via prompts, along with 10 examples and their prompts.

Google DeepMind 发布了 Genie 4 世界模型研究预览版，新增实时音频生成、更逼真的可交互世界，以及支持中途提示词编辑的实时操控功能。该模型可维持最长两分钟的场景一致性，并与 Gemini 深度集成，用于机器人仿真、游戏和影视原型制作等场景。