AI software turns simple sketches into photorealistic images in seconds

A line turns into a waterfall. A shaky circle becomes a jagged boulder. Layers of colors are suddenly a sunset. Just wait—computer drawings are about to get a whole lot more photorealistic.

Visual computing company Nvidia recently showed off a new piece of software called GauGAN (get it?) that can turn crude sketches into artfully crafted digital paintings. TechCrunch describes it as “MS Paint for the AI age,” and that’s pretty much accurate.

Using neural networks that have been trained on millions of stock landscape images, the software is able to interpret the intent of an abstract shape and turn it into a real-world scene. To be fair, the computer has a little help. Users first choose a category, whether it’s “sky,” “plant,” or “mountain;” and when they scrawl a simple shape, that sketch will transform into a much more impressive image generated by GauGAN.

What does this mean for art? What does this mean for architecture? At the moment, not much. The software is still in its development stage, but Nvidia’s computer scientists say that it could be used to create synthetic rendering images for architects on the fly. “It’s much easier to brainstorm designs with simple sketches, and this technology is able to convert sketches into highly realistic images,” said Bryan Catanzaro, vice president of applied deep learning research at NVIDIA

It does make you wonder how automated, generative software might be used by future creators to not just augment people’s work, but to make wholesale decisions about form and style.