After much teasing over the past year or two, Nvidia has finally released the Nvidia Canvas app as part of Nvidia Studio.
This awesome app lets you draw the most basic, five-year-old finger-painting style drawings and it uses AI training to automatically generate real-life imagery based on the colors, shapes, placements, etc. This is basically an AI that takes abstract shapes, applies Bob Ross’ “Happy Accidents” mantra and spits out a beautifully realistic painting out the other side. It’s mind-boggling, honestly.
I will be getting hands-on with this and updating this post with more in-depth thoughts after further testing, but I am very impressed so far.
Here’s a video by Sir Wade Neistadt using it:
A really cool possible application of this technology that doesn’t seem to be considered yet is tabletop gaming, oddly enough. The ability to just slap together a vague, abstract idea of what you want your environment pieces to look like and have them auto-generated for you could revolutionize art-making for this area. Plus, if someone created an app that uses Nvidia Canvas but also lets you do things like procedurally generate objects from sprites or pre-selected images, you could just select places you want to populate and have it auto-populate a dungeon or something. However, most of the trained images seem to be taken at a horizontal eye-level view or at a 45-degree angle, not top-down – so this won’t work for now. But soonTM maybe?