Exploring Korean AI Companion Possibilities with Live2D and Empathic Voice Interaction

Abstract

This study explores the growing trend of fostering emotional connections with AI through natural voice interactions. It presents a web-based Korean voice interaction system that integrates high-quality speech synthesis with a Live2D character. Research suggests that voice-based communication enhances emotional engagement more effectively than text-based methods, particularly in gaming and virtual reality. The system leverages OpenAI’s gpt-4o-audio-preview model to generate expressive speech while synchronized with visual elements. Designed to enhance the synergy between auditory and visual channels, it delivers a well-balanced multimodal experience. Quantitative and qualitative evaluations confirm enhanced user immersion and emotional satisfaction, supporting deeper human-AI interactions. This study provides a practical framework for advancing multimodal AI interactions, offering valuable insights for both research and real-world applications in immersive AI communication.