Nvidia releases Canvas. This app converts sketches into photo-realistic images of landscapes. Canvas works on all Nvidia RTX GPUs and will be part of Nvidia Studio, which also includes Nvidia Broadcast and Omniverse.
Nvidia releases Canvas as a free app. The program is currently in a beta phase. Canvas is based on GauGAN, a kind of AI painting tool that Nvidia developed with 5 million images as reference material. Nvidia already showed a demo of GauGAN in 2019 , which is similar to the final Canvas app.
The app lets users draw with “materials” instead of colors. For example, users can select a brush with “grass”. Brush strokes with that brush are then automatically converted to the material in question. GauGAN makes the end result as realistic as possible with the help of AI.
The beta version of Canvas has nine ‘styles’ and fifteen different materials, including water, rocks, the sky, clouds and snow. Users can save the final result as a png image. Users can also choose to export the image as a .psd file, which allows further editing of the image in Adobe Photoshop. Users with an Nvidia GeForce RTX, Quadro RTX or Titan RTX GPU can download Canvas from Wednesday. The app only works on Windows 10 with an Nvidia 445.0 driver or higher.
Furthermore, Nvidia will include with updates to the Substance 3D Collection Adobe. That collection consists of various apps that allow users to create 3D artwork . The new update also includes a Substance 3D Stager. That new app is used to place and adjust 3D objects in real time. Ray tracing is used for the lighting and shadows.
Furthermore, the Nvidia Studio drivers get support for resizable bar for GeForce RTX 30 GPUs. This allows the CPU to access the entire memory of the video card at once. Normally this is only possible with 256MB at a time. The technology is also known as Smart Access Memory, the marketing term AMD uses for this. The feature was already available in Nvidia’s Game Ready drivers. The Studio driver also gets support for dynamic boost 2.0 and GPU passthrough for virtual machines.