Interactive simulation of ultrasound imaging greatly facilitates sonography\ntraining. Although ray-tracing based methods have shown promising results,\nobtaining realistic images requires substantial modeling effort and manual\nparameter tuning. In addition, current techniques still result in a significant\nappearance gap between simulated images and real clinical scans. Herein we\nintroduce a novel content-preserving image translation framework (ConPres) to\nbridge this appearance gap, while maintaining the simulated anatomical layout.\nWe achieve this goal by leveraging both simulated images with semantic\nsegmentations and unpaired in-vivo ultrasound scans. Our framework is based on\nrecent contrastive unpaired translation techniques and we propose a\nregularization approach by learning an auxiliary segmentation-to-real image\ntranslation task, which encourages the disentanglement of content and style. In\naddition, we extend the generator to be class-conditional, which enables the\nincorporation of additional losses, in particular a cyclic consistency loss, to\nfurther improve the translation quality. Qualitative and quantitative\ncomparisons against state-of-the-art unpaired translation methods demonstrate\nthe superiority of our proposed framework.\n