In the past, "upscaling" was a dirty word in professional circles. Bicubic resampling in Photoshop gave us mushy, soft messes that looked like watercolor paintings gone wrong. You couldn't invent detail that wasn't there. That rule is dead. Technological advancements in machine learning—specifically the generative adversarial networks (GANs) and diffusion models used by Topaz Labs—have changed the physics of digital imaging. We can now add resolution while recovering (or even creating) texture, sharpness, and fidelity.... read more ›