This paper presents a diversity-based soft robotic gripper design framework combining generative design and topology optimization (TO). The framework uses Compositional Pattern-Producing Networks (CPPNs) to seed diverse initial material distributions for fine-grained TO. The approach focuses on vacuum-driven multi-material soft grippers and demonstrates several grasping modes, such as pinching and scooping, emerging without explicit prompting. Extensive automated experimentation with printed multi-material grippers confirms optimized candidates exceed the grasp strength of comparable commercial designs. The framework evaluates grip strength, durability, and robustness across 15,170 grasps. The combination of fine-grained generative design, diversity-based design processes, high-fidelity simulation, and automated experimental evaluation represents a new paradigm for bespoke soft gripper design, generalizable across numerous domains. The paper discusses the challenges of designing soft grippers for complex tasks, highlighting the limitations of existing universal grippers and the need for bespoke designs. It introduces a computational design approach that combines TO with evolutionary design to generate diverse, high-performance grippers. The framework uses a multi-material SIMP formulation for topology optimization, incorporating the Darcy method for pneumatic soft robot optimization. The cost function includes terms for tip displacement, stiffness, closure, and adaptability, with different penalty formulations to balance these objectives. The paper presents results from five sets of experiments, evaluating the performance of optimized grippers in terms of strain energy, output displacement, and energy loss. The results show significant variability in design morphology, with some designs exhibiting high displacement and low strain energy. The paper also presents experimental results on grasp performance, durability, robustness, and generality, demonstrating that optimized grippers outperform commercial designs in terms of grasp strength and robustness. The study highlights the importance of diversity in design and the potential of computational design to enable scalable, bespoke soft gripper design.This paper presents a diversity-based soft robotic gripper design framework combining generative design and topology optimization (TO). The framework uses Compositional Pattern-Producing Networks (CPPNs) to seed diverse initial material distributions for fine-grained TO. The approach focuses on vacuum-driven multi-material soft grippers and demonstrates several grasping modes, such as pinching and scooping, emerging without explicit prompting. Extensive automated experimentation with printed multi-material grippers confirms optimized candidates exceed the grasp strength of comparable commercial designs. The framework evaluates grip strength, durability, and robustness across 15,170 grasps. The combination of fine-grained generative design, diversity-based design processes, high-fidelity simulation, and automated experimental evaluation represents a new paradigm for bespoke soft gripper design, generalizable across numerous domains. The paper discusses the challenges of designing soft grippers for complex tasks, highlighting the limitations of existing universal grippers and the need for bespoke designs. It introduces a computational design approach that combines TO with evolutionary design to generate diverse, high-performance grippers. The framework uses a multi-material SIMP formulation for topology optimization, incorporating the Darcy method for pneumatic soft robot optimization. The cost function includes terms for tip displacement, stiffness, closure, and adaptability, with different penalty formulations to balance these objectives. The paper presents results from five sets of experiments, evaluating the performance of optimized grippers in terms of strain energy, output displacement, and energy loss. The results show significant variability in design morphology, with some designs exhibiting high displacement and low strain energy. The paper also presents experimental results on grasp performance, durability, robustness, and generality, demonstrating that optimized grippers outperform commercial designs in terms of grasp strength and robustness. The study highlights the importance of diversity in design and the potential of computational design to enable scalable, bespoke soft gripper design.