How to do you leverage your entire CPU/RAM/GPU?

Do you just open Topaz Video AI three times overnight, and run three exports concurrently?

I wrote a Python script to somewhat automate things. It calls normal ffmpeg to trim the source, then the TVAI enabled ffmpeg to either denoise, upscale, or inerpolate (or all three), then normal ffmpeg again to convert the processed video into reasonable H.265. It’s made in such a way that I can have it run as many instances of ffmpeg as I want. Usually I don’t do more than three. Anyway, the only thing left I need to do is join the audio and chapters back in. I’ll probably add that to the script some day. For now, it’s nice to plan out what I want done, then run the script and come back the next day to have all the different passes done.

Sounds really useful. May I have the script so I can streamline my process?

sounds amazing. care sharing the script?

It’s not very user friendly right now. Let me make it slightly better when I get some free time, then I’ll upload it.

Here it is. I explain how to use an older version of it here.
FullRun.zip (5.1 KB)

In this version I put all the paths and number of processes to do at a time in global variables near the top. I should also mention it needs ffprobe.
I’ve added a bunch of options, again they’re more specific to what I use TVAI for. There is a Theia pass and in it I give the option check the options file for the parameters (Just waiting for Theia to get fixed before I can use it). I suppose that could be useful for the Proteus pass too, but I have not needed it yet. Therefore, the parameters are hardcoded.

Anyway, good luck.

Nice! In v3 it’s not really necessary to create a cropped version, u can do that with a ffmpeg command, so that only the content part is processed by TVAI. It’s one of the features of v3 I really appreciate.
I explained it here:
https://community.topazlabs.com/t/crop-pad-for-free-performance-in-v3

I remember reading that. I have not gone through the trouble of putting in my script though. Mostly because TVAI should be optimized eventually to the point where that makes no difference. But also, the movies I’ve been processing are 4:3 and would not benefit from that.