Proteus model - how does the "estimate" parameter into FFMPEG work?

Topaz Video AI General
1

Hi,

I’m using the Proteus model (in auto mode) to enhance a large batch of old videos. I’ve written my own program which calls Video Enhance’s versions of FFMPEG, but I was just wondering what the “estimate” parameter of the tvai_up filter does, and if it’s worth tweaking?

tvai_up=model=prob-3:preblur=0:noise=0:details=0:halo=0:blur=0:compression=0:estimate=20:device=0:vram=1:instances=1

By default, it always appears to be set to 20. If I increase this to say, 100 then I notice the process takes a lot longer to start. So I’m guessing it’s something to do with the number of frames it analyzes to determine the best values for the other parameters? Perhaps some kind of rolling average window size?

Just curious in case it’s worth experimenting with. Want to make sure I get the absolute best possible settings before I set it going (will take weeks to process everything)

Thanks for any help

2 Likes
2

When I first saw that in the command, I ran a few runs with different values in it. The file size came out the same at every value I tried.

My guess is that it has something to do with relative to auto. Maybe I’ll do some more tests. I think I had copied a command to do full auto when I did my tests.

3

It would be great if there were detailed documents on the parameters of the models.