An actuator-free aberration correction using a multiple-input deep convolutional neural network is presented. It uses an additional wavefront sensor input to correct time-varying low-order aberrations. These advantages were transferred to aberration-corrected imaging velocimetry.