Graphics giant, Nvidia, has unveiled an impressive new AI-driven image creator during its GTC 2019 event with the ability to transform child-like MS Paint drawings into near photorealistic images.

It’s called GauGAN and it’s capable of some truly wild stuff. By simply drawing a crude landscape that even a 3-year-old could best, the software uses a neural network to create an image from that drawing based on what it’s learnt from a huge number of actual images.

All you have to do is select what you’d like to draw in – be it mountains, trees, water, etc – and use the mouse to draw it, just like MS Paint. You can see it in action for yourself below.

Impressive, no? The company says GauGAN produces an image unique to the input it’s given, and can even pick up on contextual changes. For example, as the video above highlights, if the ground is covered in snow, the sky should be a different colour compared to a grassy field.

But even if two different users create the same drawing with the same settings, it’ll still pump out two different images.

What you see in the video is demonstrated on Nvidia’s RDX Titan GPU, which is essentially its latest and greatest graphics processing tech. What that means is that it’s able to produce the real-time results you saw as the drawing was created, which probably won’t be possible on lower spec PCs. You could still use the software, it would just take a little longer to render the output.

Of course, GauGAN isn’t available to the public just yet, but Nvidia hopes to make it available on its AI Playground site in the future. In the meantime, the company’s white paper on the topic will be released today.

The images it produces are far from perfect but will no doubt be incredibly useful for a number of industries, including game development. Hey, it would be a pretty fun addition to MS Paint, right?

Source: Tech Crunch
Image: Nvidia