Happy Little AI Trees: Bob Ross Meets Nvidia Canvas
As a creative technologist, I’m obsessed with making technology accessible. It doesn’t have to be a scary black hole filled with code and fancy terminology. Just like Bob Ross gave millions the confidence to grab a paintbrush, tools like Nvidia Canvas make it easy to create art with just a few clicks. So what better way to test out Canvas than by doing a Bob Ross tutorial?
WTF is NVIDIA Canvas?
Canvas is one of the latest pieces of software to come out of Nvidia. It uses AI to create landscapes with just a few clicks on your screen. It works by using what’s called a GAN. This GAN was trained on a large collection of landscape photos that tell the computer how to interpret your brush strokes.
When you draw on the left panel what you’re actually doing is creating a segmentation map. That map is then used to influence where the specific features should live within the scene. Then the trained model is smart enough to take each element and blend them into a “photoreal” scene. Enough talking though! Let’s actually make something!
Marking Up the Canvas
Honestly, the most difficult part of this experiment was deciding which Bob Ross tutorial to follow. Ultimately I decided to follow the Sunset Aglow Tutorial. While watching, I noticed that Bob’s tutorial can be broken up into two core components: the sky and the landscape itself
The Sunset Sky
While Bob spent the first ten minutes applying a tasteful gradient across his canvas, I spent mine deciding on the style. Canvas offers nine styles to choose from. Each of the styles offers a different aesthetic to the final piece. After experimenting with each option, I found that style #6 was pretty spot on to what Bob had. Next up, each of the landscape features.
In-Scene Elements
Just above the style selection pan, are different “brushes” to choose from. Each one is capable of creating a different element within your scene. To create a cloud, for example, just click, drag, and watch the magic happen.
After creating the clouds I moved on to the trees, mountains, and river — click, drag, magic. Five minutes later, my masterpiece was complete.
The Final Result
Canvas is no Bob Ross, but let’s be honest, neither am I. And it would have taken me a lot longer than five minutes to create a way less polished landscape.
Another cool thing about Canvas? All the content is procedural. So if I want to see what my mountains might look like at night, I just pick a new style and the GAN does all the work.
If you think this is impressive, the future is going to blow your mind. This trend of using GANs is still in its infancy. The software is only going to get better and the tech more powerful. And it’s only a matter of time before industry-standard tools, like Photoshop, start to incorporate these features.
Final note, Nvidia Canvas does require an RTX series graphics card. Don’t have the hardware but still want to give Canvas a try? You’re in luck! Nvidia released a web-based tool that you can use right in your browser. You can check it out here. Can’t wait to see the landscapes you create!
