If you’ve ever marveled at the ability of Photoshop’s Content-Aware Fill feature to seem to magically reconstruct missing details from a photograph — or generate plausible new textures where none existed before — you’ll want to see this video overview of the next generation of similar technology being researched at Nvidia, where deep-learning techniques make CPUs and GPUs even smarter about painting pretty pictures. A research paper describing the technique in detail is available for download through the Cornell University Library.