Maybe there are more complications than I can fathom, but I just have to know why it is that frames of pure black take just as long to enhance as normal images?
Mostly I just run 780x420 to 1920x1080 and have the output set to mp4 at CRF 18. So maybe there’s some other combination where it does speed up.
I suppose I can be thinking about how it all works in the wrong way. I am thinking it should work similar to the way the png format compresses images. The size comes out smaller the less complex the image is.
In the case that the process treats all frames the same (they all get the same steps executed on them), it could benefit everyone if near solid color logic was added to bypass the AI steps on them and speed things up.