This paper proposes a new approach for compensating input delays in nonlinear systems under sampling. The proposed dynamic controller is based on an approximated predictor of the state of the system, obtained from an approximated representation of the sampled system dynamics, and ensures practical stabilization of the exact sampled system. The approximated representation can be obtained using well-known techniques (Euler, Runge-Kutta, Takagi-Sugeno, etc.), and the controller exists provided that the exact and approximated sampled dynamics verify conditions expressing their closeness.