Histopathological visualizations are a pillar of modern medicine and biological research. For example, surgical oncology relies heavily on post-operative histology to determine surgical success and guide adjunct treatments. However, current histology, featuring bright-field microscopic assessment of histochemical stained tissues exhibits several major challenges. Developing stained specimens for brightfield assessment requires laborious sample preparation, potentially delaying adjunct treatments for days or weeks. Hence, there is motivation for an alternative histopathology method mitigating the challenges of current techniques. Here, a deep-learning-based method is presented for label-free virtual histochemical staining of total-absorption photoacoustic remote sensing (TA-PARS) images of unprocessed tissue sections. The TA-PARS provides an array of contrasts (scattering, and radiative and non-radiative absorption) ideal for developing AI-based H&E colorizations. In this work, a Pix2Pix generative adversarial network (GAN) model, is applied to develop virtual H&E staining from label-free TA-PARS images. Thin sections of human skin tissue are first virtually stained with TA-PARS, then are chemically stained with H&E producing a direct one-to-one comparison. One-to-one matched virtually- and chemically- stained images exhibit high concordance validating the digital colorization of TA-PARS images against the gold-standard H&E. Results demonstrate that virtual staining of TA-PARS images can provide label-free data of similar diagnostic quality to traditional H&E. This paves the way for TA-PARS rapid slide-free histological imaging, to avoid the typical labor-intensive chemical staining procedures. In the future this may substantially reduce histological imaging times, while simultaneously reducing staining variability by removing operator dependency during histochemical staining.