In recent years, 3D cell culture models such as spheroid or organoid technologies have known important developments. Many studies have shown that 3D cultures exhibit better biomimetic properties compared to 2D cultures. These properties are important for in-vitro modeling systems, as well as for in-vivo cell therapies and tissue engineering approaches. A reliable use of 3D cellular models still requires standardized protocols with well-controlled and reproducible parameters. To address this challenge, a robust and scaffold-free approach is proposed, which relies on multi-trap acoustic levitation. This technology is successfully applied to Mesenchymal Stem Cells (MSCs) maintained in acoustic levitation over a 24-h period. During the culture, MSCs spontaneously self-organized from cell sheets to cell spheroids with a characteristic time of about 10 h. Each acoustofluidic chip could contain up to 30 spheroids in acoustic levitation and four chips could be ran in parallel, leading to the production of 120 spheroids per experiment. Various biological characterizations showed that the cells inside the spheroids were viable, maintained the expression of their cell surface markers and had a higher differentiation capacity compared to standard 2D culture conditions. These results open the path to long-time cell culture in acoustic levitation of cell sheets or spheroids for any type of cells.