3D X-ray virtual histology (XVH) techniques utilise X-ray tomography for non-invasive 3D imaging of biological samples, eliminating labour-intensive sectioning and staining required in conventional 2D histology. However, such XVH approaches yield greyscale data, lacking the biological specificity provided by histological stains in the form of selective colouring of biological tissue components. Concurrently, in emerging digital pathology, machine learning enables ‘virtual staining’ of visible light-based images, whereby histological stains may be virtually transferred onto non-stained images. Here, we bridge XVH and virtual staining approaches by using cross-modality image translation models to generate artificially stained 3D X-ray histology data from micro-CT data. In our studies on biodegradable metal implants in bone, we performed synchrotron-based micro-CT and toluidine blue histology sequentially on the same samples. Using over 50 co-registered image pairs, we trained a modified cycleGAN model suited for limited paired data, with results optimised to replicate the colours present in the histology images whilst retaining a majority of high contrast and resolution features.