Google发布世界模型 Genie 4，实时生成画面音频与对话
'%20fill-opacity='.8'%20d='M3.256%203.913a3.083%203.083%200%200%200-.003%204.397L8%2012.998l4.743-4.684a3.085%203.085%200%200%200%20.001-4.4c-.6-.593-1.4-.914-2.233-.914a3.17%203.17%200%200%200-1.91.635L8%204.087l-.601-.452A3.17%203.17%200%200%200%205.489%203c-.834%200-1.634.321-2.233.913m10.19%205.111-4.748%204.69a.996.996%200%200%201-1.397%200L2.549%209.02a4.083%204.083%200%200%201%20.004-5.82A4.17%204.17%200%200%201%205.488%202c.907%200%201.787.29%202.512.835A4.17%204.17%200%200%201%2010.51%202c1.093%200%202.146.422%202.936%201.202a4.085%204.085%200%200%201%200%205.822'%20/%3e%3c/svg%3e)
'%20stroke-opacity='.75'%20stroke-width='1.5'%20stroke-linejoin='round'/%3e%3c/svg%3e)
GoogleDeepMind发布世界模型Genie4，可实时生成画面、环境音与可交互对话，并支持用自然语言随时改动场景。模型由Gemini家族支撑，Gemini3.8Flash担任引导agent拆解多步任务。现向可信测试者开放研究预览，随后面向ProjectGenie的Ultra订阅者及GeminiAPI、AISt
参与讨论
GoogleDeepMind发布世界模型Genie4，可实时生成画面、环境音与可交互对话，并支持用自然语言随时改动场景。模型由Gemini家族支撑，Gemini3.8Flash担任引导agent拆解多步任务。现向可信测试者开放研究预览，随后面向ProjectGenie的Ultra订阅者及GeminiAPI、AISt