Nvidia's AI upscaling comes to web video

If you own a 4K monitor, watching web video on a PC is often an exercise in staring at blurry pixels. Outside of YouTube and Netflix, most streaming services cap desktop browser playback at 1080p, leaving high-end displays vastly underutilized.

At CES 2023, Nvidia announced that its AI upscaling tech is finally coming to web browsers. Starting in February, RTX Video Super Resolution will allow PCs equipped with RTX 30 and 40 series GPUs to upscale 1080p video to 4K inside Google Chrome and Microsoft Edge.

AI upscaling leaves the gaming silo

Nvidia demonstrated the feature using a 1080p clip of Apex Legends gameplay. When zoomed in, the AI upscaler visibly smooths out blocky compression artifacts on both foreground and background objects, producing a much cleaner 4K output.

However, the demonstration featured relatively slow motion. It remains unclear how the algorithm handles rapid camera pans, fast-paced action, or abrupt scene cuts without generating ghosting.

There are also notable early limitations. The tool currently only upscales from 1080p to 4K, meaning massive amounts of lower-quality 720p web content are left out for now. Browser support is restricted to Chromium-based Chrome and Edge, leaving Firefox behind.

Puzzlingly, RTX 20 series graphics cards are not supported, even though Turing GPUs feature the same fundamental Tensor core architecture that powers DLSS. Nvidia offered no explanation for cutting out the 20 series, nor did it mention whether offline local video files will eventually work.

Why it matters

For everyday desktop users, this could be an instant quality-of-life upgrade. Streaming services like Amazon Prime Video and Hulu routinely gatekeep 4K playback behind smart TVs and streaming sticks, restricting PC browsers to muddy 1080p streams. If Nvidia's tool works as advertised across any web video, it effectively bypasses those artificial platform restrictions right at the display output.

Strategically, Nvidia is expanding its edge AI moat from hardcore gaming and creative apps into everyday digital media consumption. It gives consumers another tangible reason to keep paying the RTX hardware tax.

We think shifting the upscaling workload to client-side hardware is a clever move that saves streaming platforms bandwidth. The big question mark is efficiency. If running real-time neural upscaling spins up GPU fans and spikes power draw just to watch a YouTube video, users might think twice before leaving it on 24/7.