real-esrgan for video is upscaling size but not improving quality

[tacitus-git]

I am running the script from an anaconda environment with the argument: python inference_realesrgan_video.py -n RealESRGAN_x4plus -i inputs\video\test.mp4 -s 2
The script appears to work. All is completed without error. However, although the video is indeed upscaled in size by 2x, the quality remains the same. This video is live action. What could be wrong. Any ideas?
Cuda is installed and running fine on a rtx 3060 for the processing.