Duplicate AI Upscaling (ex. NVIDIA RTX Super-Resolution)

[MagicDoubleM 121]

In contrary to transcoding for getting data rates down and allowing clients to play media that doesn't have compatible codecs, I propose a similar approach in getting the max out of low quality media.

So like for example, we have a SD-Video, but also clients with fast connection and capable of playing back HD and 4k. Some of us use beefier PCs and fast GPUs that are capable of doing some fancy upscaling and post-processing, even in real-time. Would be fun to have the server just taking care of it, right?

I don't expect wonders, but for example something in the realm of NVIDIA Shield's upscaling seems to be possible, because, when I play with the tool from NVIDIAs Maxine SDK I get similar results. They might not make a huge difference, but in motion it's a visible and an interesting option, depending on the content.

I haven't found out if there is a ffmpeg-patch, which would allow to leverage that NVIDIA-implementation easily, somewhere to be discovered yet.

[Luke 42078]

Hi, there's already an open request for this or something functionally equivalent. Please join in and contribute to the existing discussion at: