A synthetic fingerprint generation method is presented and implemented in this paper. A new combination model for the orientation field of fingerprint is applied to generate more realistic fingerprint direction map, and a new expression of the fingerprint density map is presented. Then a modified Gabor filter is used for ridge pattern generation. This method consists of two main processes: First, a master-fingerprint image is generated after direction map generation, density map generation and ridge pattern generation. Then a realistic synthetic fingerprint is obtained through a series of steps on the master-fingerprint, including adding scratches, displacement of ridges, erosion or dilation of ridges, distortion, noising and rendering, global displacement and rotation, changing contrast and adding background noise. The synthetic fingerprint generation platform implemented with this method has been applied and verified in Biometrics Verification Competition 2004. Finally, the numerical experimental results approve the validity and robustness of the method.