Abstract Soft robotic grippers face significant challenges in perceiving and manipulating objects with varying stiffness and size, particularly in unstructured environments. This paper introduces a novel tactile sensing system integrated into a soft robotic finger, which uses a deformable silicone elastomer substrate as a physical reservoir coupled with a pressure sensor to independently estimate object stiffness and size. The system leverages the substrate’s mechanical compliance to encode distinct responses related to object characteristics, enabled by an innovative design that integrates material properties with advanced signal processing. A regression model was trained on a compact dataset of four samples using steady-state voltage responses, achieving high predictive accuracy of 99% for predicting the size of a cube. Experiment results validate the system’s reliability in differentiating object properties with limited data. By embedding the necessary computation to estimate stiffness and size on the finger material properties, this approach has the potential to enhance perception in soft grippers handling delicate or irregular objects, with broad implications for agriculture, manufacturing, and medical robotics. Future efforts will expand datasets and incorporate multi-modal sensors to enhance robustness and versatility.