The images of various flower species in the Oxford Flower dataset are natural and complex, providing a unique context compared to conventional image processing tasks. This study conducts experiments and evaluates the capabilities of the Deep Convolutional Generative Adversarial Network (DCGAN) model, combined with BERT to enhance the understanding of context and attributes of the flowers through textual descriptions. The model employs convolutional neural networks for the discriminator and deconvolutional neural networks for the generator to create images of the flower species. The experimental dataset is the Oxford Flower -102 dataset. The results indicate that the DCGAN combined with BERT demonstrates effective applicability and potential in multimedia data generation tasks, thanks to its ability to integrate contextual information from textual descriptions into the image generation process.