Aiming at the needs of reconstructing the structure of calligraphic seal cutting strokes and virtual display, this study designs a GAN technique that integrates three models, namely, “WGAN, DCGAN and CGAN”. The Cycle GAN model is used to obtain the mapping relationship between learning and style migration by utilizing its cyclic consistency loss. Adaptive pre-morphing technique is introduced to process the input image to capture the outline information and morphological features of calligraphic seal carvings, and a Generative Adversarial Network-based Generative Model for Structural Reconstruction of Calligraphic Fonts (CRA-GAN) is proposed. Meanwhile, an online virtual display system is designed to provide users with a good sense of experience in the virtual display of calligraphy. The results show that the CRA-GAN model can better capture the details and global information of the fonts, and its recognition rate of the eight calligraphic fonts ranges from 90.42% to 97.38%, and the MOS rating value of the text image is > 8.5 points, and its recognition results are in line with the observation characteristics of the human eye for calligraphic images. The FID calculation result of the CRA-GAN method ( 204.361) of the CRA-GAN method is much lower than that of other methods, which obviously improves the diversity and visual quality of the generated calligraphic fonts. This paper evaluates the user’s experience of the system from five aspects: narrative experience, emotional experience, sensory experience, cognitive experience and interactive experience, and calculates that the final score of the system is in the range of 80-100, which indicates that the user’s satisfaction is very high after actually experiencing the virtual display system.