Twisted string actuators (TSAs) generate high tendon-based (muscle-like) linear actuation and forces and have wide operational bandwidth. For these properties, TSAs are well-suited for advanced robotic and mechatronic applications, though their inclusion in soft robots has been limited. We recently employed TSAs to drive an anthropomorphic soft gripper which successfully demonstrated dexterous grasping and manipulation. While the gripper showed superior qualitative performance, the lack of modeling and control strategies limited its realization in a fully autonomous system that could generalize well to new grasping behaviors. As a first step toward this objective, this paper presents modeling strategies for a monolithic soft silicone finger driven by a TSA. A physics-based kinematic model is developed to predict the bending and velocity kinematics of the finger as a function of the TSA's twisting angle. An average error of 2.06° was obtained for the kinematic bending model across different actuation behaviors and an average error of 2.95°/s was measured for the angular velocity model.