Fluid Body is an experimental, body-driven instrument with which participants can use a compact classical ballet arm-posture vocabulary to play a digital Guzheng. The system implements an interactive machine-learning mapping that combines a posture classifier to label each recognized posture, with a regression model for continuous control of granular sound synthesis parameters. Six ballet arm positions are associated with six edited Guzheng sound events, enabling participants to trigger distinct phrase regions through posture changes while continuously modulating the timbre in real time. We use intercultural co-adaptation in a dynamic, operational way. Rather than claiming literal translation between ballet and Guzheng traditions, the system creates a performable correspondence space where a codified movement vocabulary can be used to explore a traditional instrument’s playing through an adaptive mapping. Fluid Body therefore contributes a concrete hybrid mapping design for embodied sonification in an intercultural context, demonstrating how machine learning can serve as a creative translator between different cultural artistic expressions.