The application of handwriting analysis in the health field for early detection and diagnosis is limited by a lack of data, which presents a significant challenge for the implementation of deep learning-based models. To address this issue, numerous studies have focused on generating synthetic data. Current methods for generating synthetic handwriting from offline images, including generative models and geometrical techniques, fail to account for the kinematics of handwriting movements, which is critical for achieving more human-like results. To address this limitation, we propose a novel, human-like approach for the synthetic generation of handwriting or drawing from offline images. This method creates new samples by incorporating a trajectory recovery algorithm and a human-like time law generator, as well as the extraction and manipulation of kinematic parameters. The evaluation of the proposed method from both visual and kinematic perspectives demonstrates its potential applicability across a wide range of devices and handwriting styles.