Seismic full waveform inversion (FWI) is a technique to build a high-resolution velocity model of the subsurface by iteratively minimizing the misfit between recorded and synthesized seismic data. However, classical FWI driven by gradient-based local optimization is vulnerable to local minima caused by lack of low-frequency components and an accurate initial model. Although global optimization methods such as genetic algorithm (GA) are less affected by the presence of local minima, those methods are affected by "curse of dimensionality." This results in low-resolution model less than optimum solution. Therefore, we propose an FWI method based on genetic local search algorithm with hybrid-grid scheme (HGLS-FWI). This method combines GA with coarse grid as a global search and gradient-based optimization with fine grid as a local search to directly deliver high-resolution model, while reducing the risk to be trapped in a local minimum. Our experimental results demonstrated that the proposed method reduced the average velocity estimation error by 62 compared with a classical gradient-based FWI on the condition that neither low-frequency components nor an accurate initial model was available. © 2019 IEEE.