This paper has addressed the fixed-time synchronization of neural networks. Aiming to realize fixed-time synchronization of driven-response system, two different structures of controllers are established based on finite-time theorem. Different from the existing methods, the powers of controllers’ fractional term can vary with the value of states, which lead controller to be simpler and can be tolerant to disturbances or uncertainties. Moreover, corresponding conditions to select controllers’ parameters are given out. Finally, a numerical example with simulations is employed to validate the effectiveness of the theoretical results.