DeepMind Launches Genie 3, a Text-to-3D Interactive World Model
Briefly

Genie 3 by DeepMind is an advanced generative framework for creating interactive 3D environments from text prompts. It operates at approximately 24 frames per second in 720p resolution, facilitating continuous interaction for several minutes. A notable improvement is its object permanence, where changes to the environment persist over time. Genie 3 acts as both a unique content creation tool and a simulation platform, offering varied settings for applications in robotics and AI. Its ability to generate content procedurally sets it apart from other models in the field.
DeepMind's Genie 3 generates interactive 3D environments from text, enabling persistent object manipulation and real-time interaction, distinguishing it from other AI models.
Genie 3 operates at 24 frames per second in 720p, supports continuous navigation for minutes, and allows for various environments based on user prompts.
The model's core innovation is object permanence, ensuring environmental changes remain over time, which enhances training scenarios in robotics and AI applications.
Combining content creation and simulation, Genie 3 provides rapid prototyping capabilities for autonomous agents to develop generalizable skills across dynamic worlds.
Read at InfoQ
[
|
]