Nvidia's RTX Video HDR promises to bring AI-generated HDR to your simple SDR streams this month

pcgamer.com:

Alongside the RTX 40-series Super cards and RTX Remix launch at this year's CES 2024 event, Nvidia has also announced another AI feature heading to its GeForce RTX drivers later this month. Joining RTX Video Super Resolution, which upscales low resolution video streams in a browser, is RTX Video HDR: A deep learning algorithm that will automatically convert an SDR video stream into a HDR one.

Nvidia claims that a lot of content streamed off the likes of YouTube is low resolution, low bandwidth, and standard dynamic range content. And if you're watching it on a decent 1440p or 4K monitor, then it just won't look very good. This is why Nvidia developed RTX VSR (video super resolution), a sort of DLSS for video streams, to upscale something like 540p clips to 1080p or higher.

That addresses the low-resolution issue and now Nvidia is tackling another one. Lots of the latest gaming monitors are HDR-compatible, which means it can use more bits per pixel, to create a wider range of colours (aka high dynamic range, HDR). However, the kind of video content that Nvidia wants to improve with RTX VSR is also in standard 8-bit (aka SDR).

Cue RTX Video HDR to the rescue! We don't have the full details on exactly how it works right now but I'm guessing that it works just like DLSS and RTX VSR do: A deep learning model is trained on what an HDR version of a video should look like and then your GeForce RTX graphics card runs a neural network, via its Tensor cores, to convert the video stream into a HDR one.

While RTX Video HDR will work on any GeForce RTX card (i.e. 20, 30, and 40-series), you will need a monitor that's HDR10 compliant. That's one of the most common HDR formats around and pretty much any monitor that claims to be