This study investigates the feasibility of deploying an AI-based fitness platform
within a metaverse environment. The proposed platform integrates three core AI
technologies—pose estimation, facial emotion recognition, and large language models
(LLMs)—to deliver personalized feedback and immersive exercise experiences. The pose
estimation component tracks user movements in real time, providing corrective guidance
to ensure accurate performance. The facial emotion recognition module analyzes users’
affective states to deliver motivational prompts and emotional support. Meanwhile, the
LLM enables natural, conversational coaching and interaction, thereby enhancing user
engagement and exercise adherence. By combining these AI capabilities, this research
proposes a next-generation, metaverse-based fitness experience. Experimental results
indicate that the platform can positively influence both user satisfaction and exercise
efficiency