Full-waveform inversion is an important and widely used method to reconstruct subsurface velocity images. Waveform inversion is a typical nonlinear and ill-posed inverse problem. Existing physics-driven computational methods for solving waveform inversion suffer from the cycle-skipping and local-minima issues, and do not mention that solving waveform inversion is computationally expensive. In recent years, data-driven methods become a promising way to solve the waveform-inversion problem. However, most deep-learning frameworks suffer from the generalization and overfitting issue. In this article, we developed a real-time data-driven technique and we call it VelocityGAN, to reconstruct accurately the subsurface velocities. Our VelocityGAN is built on a generative adversarial network (GAN) and trained end to end to learn a mapping function from the raw seismic waveform data to the velocity image. Different from other encoder-decoder-based data-driven seismic waveform-inversion approaches, our VelocityGAN learns regularization from data and further imposes the regularization to the generator so that inversion accuracy is improved. We further develop a transfer-learning strategy based on VelocityGAN to alleviate the generalization issue. A series of experiments is conducted on the synthetic seismic reflection data to evaluate the effectiveness, efficiency, and generalization of VelocityGAN. We not only compare it with the existing physics-driven approaches and data-driven frameworks but also conduct several transfer-learning experiments. The experimental results show that VelocityGAN achieves the state-of-the-art performance among the baselines and can improve the generalization results to some extent.