High-rate loading induced cavitation of soft materials, including brain and muscle, has increasingly been studied due to its importance in a variety of biomedical applications, such as potential impact-related traumatic brain injury and high-intensity focused ultrasound therapy. The size of cavitation nuclei or defects is crucial to the onset of cavitation. However, it remains challenging to introduce gaseous defects with controllable size into soft materials and to understand the underlying mechanism of impact-bubble interaction in soft materials. Here, we set up a drop-tower system to perform impact loading tests on gelatin samples with controllable monodispersed microbubbles, allowing us to mimic the general cavitation behavior of human tissues. The microbubbles are inserted into the soft matrix as cavitation nuclei by taking advantage of microfluidic flow focusing. A high-speed camera paired with a signal acquisition system is utilized to determine the critical pressure at which cavitation initiates for different defect sizes. Meanwhile, we propose a theoretical model based on surface tension to predict the time-dependent spherically symmetric cavitation of a pre-existing gas bubble in elastomers under high-speed loading. Our experiments and theoretical modeling demonstrate the strong effect of gas-filled defects on cavitation and indicate the necessity and significance of controlling defect size in mimicking targeted organs using soft gels. Our work provides a promising route to predict cavitation-induced injury in tissues with variable materials stiffness and initial defect size.