We explore the utilization of artificial intelligence (AI) generative models for creating high-resolution airborne Synthetic Aperture Radar (SAR) images. Our methodology involves the use of a text-conditioned latent diffusion architecture to train a generative model. We use a database of high-resolution SAR images obtained from the SETHI sensor at ONERA for training purposes. This model is capable of generating synthetic images based on textual prompts provided by users. Additionally, we illustrate the model's versatility for various applications, such as generating SAR images from handdrawn sketches.