This research report investigates using generative AI to create 2D and 3D models for the Holobox, a system providing immersive visual experiences. The report explores several AI tools, analyzing their advantages and disadvantages for real-time avatar and virtual human generation. A prototype is proposed, focusing on voice-driven 2D visual creation using Cat-4D, Runway Realtime-Gen, and Meshy.ai, with potential future expansion to 3D models. The research concludes that combining these tools could significantly improve accessibility and real-time capabilities in creating such visuals. Further steps include exploring voice-driven 3D model generation and integrating camera-based expression control.