A method for simulating strong ground motion for a large earthquake based on synthetic Green's function is presented. We use the synthetic motions of a small event as Green's functions instead of observed records of small events. Ground motions from small events are calculated using a hybrid scheme combining deterministic and stochastic approaches. The long-period motions from the small events are deterministically calculated using the 3D finite-difference method, whereas the high-frequency motions from them are stochastically simulated using Boore's method. The small-event motions are synthesized summing the long-period and short-period motions after passing them through a pair of matched filters to follow the omega-squared source model. We call the resultant time series “hybrid Green's functions” (HGF). Ground motions from a large earthquake are simulated by following the empirical Green's function (EGF) method. We demonstrate the effectiveness of the method at simulating ground motion from the 1995 Hyogo-ken Nanbu earthquake (Mw 6.9).