HD vs 4K for YouTube Uploading

[dellsam34]

We all know that uploading SD video files to YT has been a joke for quite some time now, This is why I started resizing to HD before uploading, However we are at it again with another similar scenario where even HD is not important for YT algorithms anymore, 4K files will get special treatment of VP9, but I think there is more to it than just VP9, It looks like even AVC1 HD files are more compressed than it use to be, But let's assume this is not true and just focus on 4K VP9 vs HD AVC1.

I keep saying 4K and I know it's the wrong name as 4K has a slightly different resolution, It should be called UHD, but just calling it 4K for the sake of public understanding.

I've uploaded two identical captures, One was de-interlaced and resized to 1440x1080 and uploaded as such (lossless HuffYUV), The second link is the one that was also de-interlaced and resized to 2440x2160 and uploaded as such as well, Please feel free to discuss and analyse, any opinion is greatly appreciated.:

HD upload:
https://youtu.be/8f57a_ouoYo

4K upload:
https://youtu.be/SIwhelThKoE

I want to point out that I'm aware that all this resizing means nothing to the original SD master files (even if they are broadcast quality) when played back locally or shared in other ways besides the streaming platforms, I'm just focusing on the mainstream sharing platforms, specifically YouTube and discuss ways to improve the visual quality of a shared video file from when it leaves the hard drive to when it reaches the other end of the screen. I don't want people to think I'm trying to make VHS look like IMAX, I'm not. So please avoid this discussion.

Now uploading resized 4k files in long lengths is challenging in terms of file size, So an encode is a must, I would also like to discuss ways of how to encode these 4k files before uploading, Script, NLE, Hybrid ...etc, Anything in your mind.

[lollo]

For what is worth and with a quick check, I see that the HD and 4K video looks very similar in term of image quality (fixed frame). Maybe the 4K frame is a little bit sharper. I had to resize the HD frame to 4K with nnedi3_rpow2(rfactor=2, nns=4, qual=2): https://imgsli.com/MzA5MTk0

You could also try the opposite, downscaling the 4K to HD and compare.

In term of motion quality, I do not see any major difference.

For uploading, if you remember our old discussions, we found that, although HuffYUV is referable, if you compress with h264 (HD) or h265 (4K) using a low CRF there is not much penalty in quality (YT further compresses anyhow).
Be careful with ColorMatrix parameter specific for HD and 4K (Rec.709 and Rec.2020)

[Alwyn]

This rule has applied for at least two years: Any video 1440 or higher gets VP9. Below 1440 gets H264.

[dellsam34]

If you download the samples it may not be much of a difference but in terms of direct playback out of the platform I see a lot of compression artifacts for the HD version such as holding parts of the image still (not sure what they call this compression scheme), blocking, glitching. I have fiber internet so I know my network is not a bottleneck for playback. And yes the HD version and the HD from 4K are very similar but that's not the point here.

At the "This is Michael Bolton" segment is where you see the difference in compression clearly.

[lollo]

I see a lot of compression artifacts for the HD version such as holding parts of the image still

I don't (there is some "pause" in the segment you mentioned, but is there for 4K version as well, so I suppose is in the master)

[dellsam34]

Much worse in the HD version, not in the capture obviously, it's lossless.

[lollo]

Not seeing it sorry, let see what others say...

[poisondeathray]

There are problems with both versions, both AVC and VP9, at 1080 and UHD

It looks to me like there were originally 23.976p content telecined parts that were left at 59.94p with duplicates. This causes artifacts and motion disturbance issues when YT re-encodes - duplicates are not encoded true duplicates but instead there are motion estimation errors -it looks like partial movement and blocks into the next frame, when it should have been a true duplicate. ( QTGMC can contribute to those types errors too, you get a slight temporal blend on a duplicate - sort of a wobbly frame - compared to a true field match and weave and clean duplicate. You often see this on animated content with duplicates when someone uses . YT will just intensify those types of errors)

If you go back to the "HyperDeckS2.mov" from the other thread , there are problems in the source to begin with. Parts of it like "23.976p" content, but there are also drops and duplicates - e.g. the bus scene on the road has jumps and duplicates . Not sure if that's in the source or because of the capture. The sailboat, MB outside the bus, and indoor scene at the end are "23.976p" content . If the main concert footage is not interlaced 29.97 content, I probably wouldn't upload 59.94p

[davexnet]

Perhaps download the equivalent version from both encodes and look at it on the PC,
at least that way you know what you're comparing - for example

[dellsam34]

PDR, Yes the original VHS has mixed contents 23.79 and 29.97 fps just by pausing the tape and playing frame by frame, Since it's mixed there is nothing I can do about it, So I de-interlaced everything to 59.94p, so yes the compression algorithm may cause parts of the frame to stay stationary across few frames because of this, But there is still other quality differences, What I'm trying to get to is, is it worth it uploading 4K or just stick with HD?

[dellsam34]

Perhaps download the equivalent version from both encodes and look at it on the PC,
at least that way you know what you're comparing - for example

That's not the point though, The goal is to compare how the original SD version looks across the screen with the two processes HD and 4K, At the end of the day they are both 704x480 or less, But we're trying to see how it comes out through both compression schemes, Imagine an artist drawing the same painting on two types of canvases, One is a coarse fabric and the other is fine fabric but both fabrics have different drawbacks to the paint used and the artist is trying to decide which one looks better after the job is done, fine fabric may look smooth but it creates bleed, coarse is rough but handles the paint better, I know it's a hypothetical example but hope it explains my point.

[davexnet]

That's fair. Did you at least turn on stats for nerds so you can see which version you're getting?
My wide screen monitor developed problems so I'm using an old, square, 1280x1024 for now - I can't even look at the higher
res encodes.

[poisondeathray]

The UHD version is better quality wise. That doesn't necesarily mean a viewer gets the UHD or VP9 version streamed to them.

Some people have problems playing VP9 smoothly, others have problems playing AVC - often there are GPU/driver/browser configuration issues .

If you don't mind the longer encodes/uploads - I think it's worth it to try to get VP9 treatment on YT - many tests show it's better.

[Alwyn]

If your video qualifies for VP9 encoding (post #3 and possibly other criteria such as channel views/subscribers), all resolutions will be streamed in VP9.