Google Genie 4 加入实时操控、原生音频与 Gemini 推理
Google 的 Genie 4 整合 Gemini 系列能力，以 Gemini Omni 支撑模拟场景的外观与行为，并由其他 Gemini 模型维持连贯性和连续性。新版本支持用文本提示实时改变或添加场景内容，提供环境声、音乐和交互对话等原生音频，并加入更真实的视觉细节、物理表现和自然人物动作。
Genie 4 的更新把实时操控、原生音频与 Gemini 推理结合在同一模拟环境中，展示了交互式世界模型的能力构成。
Genie 4 incorporates and builds on capabilities from the Gemini family: Gemini Omni anchors simulations in real-world appearance and behavior, while other Gemini models provide the reasoning to maintain coherence and continuity.
Here's what's new:
✨ Live steering: Use text prompts to change or add things in the world
🔊 Native audio: Hear ambient soundscapes, music, and interactive dialogue
🌊 Unprecedented realism: Experience a leap in visual details, from realistic physics to natural human movement
🧠 Gemini reasoning: Interprets prompts, tracks world state to remember what just happened, and plans multi-step events so scenes stay more coherent
来源：Google · x.com