Thousands of extrasolar planets and candidates have now been detected, but almost all through indirect methods, such as transit photometry or radial velocity. Though statistically powerful, these techniques provide in most cases just a basic measurement of an object’s size and orbital parameters, and are biased towards small separations. Direct imaging, by contrast, is most sensitive to planets in wide orbits (>5 AU); and if a planet’s light can be seen, it can be characterized spectroscopically. Currently this is only practical for young (below a few hundred million years) self-luminous massive (>1 Jupiter mass) planets. Although this has been done for only a dozen systems, each provides insights into the atmospheric structure and evolutionary history of such systems. This sample is expanding with new facilities such as the Gemini Planet Imager (GPI), a dedicated high-contrast adaptive optics instrument on the Gemini South telescope. Reaching contrast levels of 10-6 , GPI has reported its first planet discovery, 51 Eridani b, in 2015. This planet is sufficiently young, low-mass, and cool that it displays strong atmospheric methane features, and its luminosity likely retains the memory of its formation. I will discuss GPI, the discovery of 51 Eri b, its properties, and the emerging statistical picture of the frequency of wide-orbit planets. I will also briefly discuss future prospects for facilities such as WFIRST, HABEX, LUVOIR, and TMT, and the ultimate path to direct imaging of Earthlike planets.