UnrealZoo is a suite of large-scale, photorealistic virtual environments designed to enhance the development of general visual intelligence, particularly for embodied AI. Leveraging the Unreal Engine and UnrealCV, these environments provide a scalable, controllable, and diverse play ground for AI agents, helping to speed up advancements in embodied artificial intelligence. The use of high-fidelity graphics and physics simulations allows for the study of complex interactive visual tasks in open worlds. We outline the technical framework of UnrealCV+, emphasizing its integration with machine learning pipelines and its utility in generating synthetic data for AI research. We demonstrate the applications of the virtual environments in developing the visual intelligence for embodied agents. This project concludes with a perspective on the future integration of multi-modal inputs and the role of UnrealCV+ in shaping the next generation of AI systems with advanced visual intelligence.