Nvidia's new AI tool to breathe new life into streaming video

The technology works with RTX 30 series and 40 series cards (Image: Nvidia)

Nvidia has introduced an Artificial Intelligence (AI)-based image upscaling tool for streaming video.

Meant for the company's GeForce RTX 30 and 40 series graphic cards, RTX Video Super Resolution uses the GPU to upscale lower-resolution online video to 4K.

Also Read | HP Omen 17 gaming laptop launched with 13th Gen Intel Core i9 CPU, Nvidia RTX 4080 GPU

Nvidia said that the tool will also remove any compression artifacts in the video. Online video is typically highly compressed to improve bandwidth usage, but that comes at the cost of quality.

Heavy compression can mean the video will be blocky and lose sharpness and clarity. Super Resolution will use its AI-based algorithms to fix blockiness, ringing, banding, and washouts from loss of detail.

AI upscaling of an image is achieved by putting low-resolution sources such as video, through an upscaler running on an AI model. The model then predicts what the image would look like in high resolution, and produces the results.

Also Read | As chatbots boom, Nvidia sales outlook beats Wall Street expectations

This calculation happens in real-time and uses the power of Nvidia's Tensor Cores on the GPU to process the information. The result is superimposed with an image derived from traditional non-AI-based upscaling to correct errors and increase the sharpness.

Nvidia says it has trained its deep learning AI network on a wide range of content with various levels of compression. Once the model has learned what these artifacts look like, they are used as references in real-time scenarios.

The technology works with RTX 30 series and 40 series cards and will upscale any streaming content running in Google's Chrome browser or Microsoft's Edge browser.