Author’s note: This publication is part of the AI Interpreted collection, which elucidates AI by enhancing the accessibility of the technology, and presents novel hardware, software, tools, and enhancements for RTX PC users.
The AI landscape has been revolutionized by generative models — notably seen in widely used applications like ChatGPT and Stable Diffusion.
This transformational journey was sparked by foundational AI models and generative adversarial networks (GANs), igniting a surge in productivity and innovation.
One such groundbreaking model is NVIDIA’s GauGAN, which powers the NVIDIA Canvas app by leveraging artificial intelligence to convert rough sketches into photorealistic masterpieces.
Genesis of the Evolution
GANs are sophisticated deep learning models that consist of two complementary neural networks: a generator and a discriminator.
These neural networks engage in an adversarial competition. The generator endeavors to produce realistic, lifelike images, while the discriminator strives to differentiate between authentic and generated content. Through continuous adversarial rivalry, GANs enhance their ability to generate realistic outputs.
GANs excel at deciphering intricate data patterns and delivering top-notch outcomes. Their applications span across image synthesis, style transfer, data augmentation, and image-to-image translation.
NVIDIA’s GauGAN, named after the post-Impressionist painter Paul Gauguin, is an AI showcase for photorealistic image creation. Originated by NVIDIA Research, this model directly led to the development of the NVIDIA Canvas app — available for exploration at no cost via the NVIDIA AI Playground.
GauGAN has garnered immense popularity since its premiere at NVIDIA GTC in 2019 — embraced by art educators, creative agencies, museums, and a vast online audience.
Transforming Sketches into Scenic Artistry
Powered by GauGAN and local NVIDIA RTX GPUs, NVIDIA Canvas employs AI to convert simple brushstrokes into realistic landscapes, showcasing instantaneous results.
Users commence by sketching basic lines and shapes with a palette encompassing real-world elements such as grass or clouds, denoted as “materials” within the app.
The AI model then generates the enhanced image in real time on the other part of the screen. For instance, a few triangular shapes drawn with the “mountain” material will manifest as a striking, photorealistic mountain range. Alternatively, users can opt for the “cloud” material and swiftly transition environments from bright to overcast with a few mouse clicks.
The creative possibilities are boundless — sketching a pond will reflect other elements in the image, like trees and rocks, in the water. Altering the material from snow to grass shifts the scene from a cozy winter setting to a tropical paradise.
Canvas features a Panorama mode that empowers artists to craft 360-degree images for integration into 3D applications. YouTuber Greenskull AI showcased the Panorama mode by illustrating an ocean cove, subsequently importing it into Unreal Engine 5.
Get the NVIDIA Canvas app to commence your journey.
Delve into NVIDIA Broadcast, another AI-driven content creation app that converts any space into a personal studio. Broadcast is freely accessible to RTX GPU owners.
Generative AI is reshaping gaming, virtual meetings, and interactive encounters of all types. Stay informed about current trends and upcoming developments by subscribing to the AI Interpreted newsletter.
