1080p -> 4K upscaling: How to best achieve smaller file sizes?

What I did to try a few tests was to use the trim feature to cut down my video to around 15 seconds or so. That way I can do multiple exports with different settings and compare the output size and quality easily. As a general rule of thumb a couple easy ways to really drop file size is:

  1. Remove noise! Noise reduces the ability of the encoder to highly compress each frame of video because it’s dealing with millions of tiny flecks of detail that interrupt larger blocks of common colours that could easily be compressed without (much) loss of quality.
  2. Reduce your bitrate! Experiment, everyone has different tolerances, but you’d be surprised at how low you can reduce your bitrate and still retain great quality.

For the purposes of this test, I trimmed down an old animated show to just a handful of seconds. Then I set Topaz Video AI to use your settings. Upscale to 4k. Proteus. Auto parameters. h265, mp4, auto bitrate.

Then I took the same clip and bumped up the Auto parameters by further really reducing noise and deblurring and cranking DOWN the bitrate to just 8Mbits/sec.

The result??? Original settings that tried to match yours, was a 347.2MB file, whereas my second test by radically reducing noise, reversing compression artifacts (which is really just more noise), doing some deblur and then setting the bitrate to 8Mb/sec … aaaaaaaand … 17.8MB!!! :astonished:

So right within Video AI itself, by applying filter choices and settings to reduce noise and just overall cleanup image quality you can REALLY bring down file sizes compared to just going auto. :+1:

Here is a side-by-side of the final output of my test exports. LEFT = auto settings, RIGHT = DENOISE, DEBLUR, 8Mb/sec :smiley:

5 Likes