That's the goal of this repository but the code base doesn't support video atm. You could theoretically use OpenCV to extract video frames and upsample them on

That's the goal of this repository but the code base doesn't support video atm. You could theoretically use OpenCV to extract video frames and upsample them one by one through this model in real time (384 -> 1024 upsampling runs at 30fps).