Generative AI has lowered the barriers to music creation, providing valuable support for inspiration and realization. However, the generated outcomes often deviate from the creator's expectations, especially when the prompts are imprecise. This study integrates Tangible User Interface (TUI) with generative AI to explore how TUI can assist creators in visualizing and adjusting musical elements, thereby narrowing the gap between the outcomes and their expectations. We invited 10 participants aged 18–35 with varying levels of musical experience to test the system, conducting evaluations on subjective musical perception, interface usability, and semi-structured interviews. The results show that participants were able to generate music aligned with the task descriptions using two types of samples and gained a deeper understanding of their preferences for system operation and interface design.