We propose a multi-channel dynamic matching full-waveform inversion (DMFWI) for a high-resolution velocity-model update, which focuses on solving kinematic difference between input data and synthetic data. With the data residual calculated in localized windows in time and space, DMFWI provides a robust velocity-model update using the total energy in the data including both diving wave and reflections. The applications to a streamer dataset and a sparse node dataset with ultralong offsets in Gulf of Mexico shows its capability to resolve large velocity errors and give significant.