Video AI 3.0.5 - Image dimensions

I’m busy upscaling some old Video-8 footage (image ratio 708x572) to FHD.
I use Artemis as the AI model and set the video input condition on “low, noisy”.

The preview inside Topaz looks perfect! (see first image in attachment)
Only the resulting video becomes a “pillar” when I choose “Letterbox/pillar” on “video out” or the resulting video contains only a part of the original image (ratio 4:3). (second attachment)

Where can I tell the tool to have the same output image ratio as the original?

Att 1 - Inside Video AI

Att 2 - Pillarbox or Crop

And I found the answer also: the ratio of the source is not the one of the output.
So, if I define e.g. the output via the height = 1080 (the width = 1336 automatically) than it works!

There’s a difference between SAR (Source Aspect Ratio) and DAR (Display Aspect Ratio). Never do anything with the former: it’s essentially only used for internal usage/storage purposes. DAR is ultimately the only thing that matters, and one of the things VEAI actually gets right. :crazy_face: