Sudden Loss in Video Quality (YouTube)

[midiout]

I made a thread a few days ago based on this situation, but I think redirecting the focus of my problem might lead to a better discussion or solution. I'm not trying to overload the forums with my b.s. but I don't know where else to turn. Reading posts here everyone seems very knowledgeable on this subject.

So I started a YouTube adventure a a few weeks ago. Everything was going as planned, and the videos were resulting as I expected. Now I've hit this random brick wall, my video quality has suddenly plummeted. I've spent the last week trying to figure out what's gone wrong but I've had no success.

I've been recording my footage in Dxtory, using the Lagarith codec, at 1080p/30fps. I've been editing in Vegas, rendering as Windows Media Video V11, QVBR, 86% Quality. This was working perfectly for me for the first few videos, no matter what I did the video's were consistent and held the same level of quality.

Here's an exmaple of my old uploads:

https://youtu.be/17IyANQp0AU#t=08s (00:08 time stamp)

https://forum.videohelp.com/images/imgfiles/RbB4ix4.jpg

I was trying various other codecs for Dxtory, different render settings, I tried out premier and handbreak. This was all in an attempt to further improve on my quality. But now I've realized that something's gone wrong, and when I render my old project, with the same settings, it looks much worse.

Here's what it looks like now:

https://youtu.be/erb8kgZSXK8#t=08s (00:08 time stamp)

https://forum.videohelp.com/images/imgfiles/7k6DclA.jpg

I assumed I changed some settings I'm unaware of, or maybe that installing other codecs screwed something up, so I went as far as reformatting my computer. It hasn't helped. The same project file, source footage, project setting, and render settings are putting out much lower quality videos. I was told that it's possible YouTube has changed something with how they're compressing videos, but no one else seems to have encountered this sudden quality loss.

Any ideas? I'm assuming I've screwed something up in my Vegas settings, or possibly other related settings on my computer. I'm at a loss here.

tl;dr Videos look noticeable worse with the same settings. Super compressed gif compairson. I used to be able to see his eyes, now I can't even tell that he has a face. To be clear this isn't the same video re-uploaded, it's been re-rendered with the same settings and then re-uploaded.

[midiout]

Found the answer to this problem. It's the different ways YouTube encodes. The video with the improved quality uses VP9 where as the lesser quality videos were using various other things. I spoke with a YouTube employee and he explained to me how VP9 is somewhat random, and he requested my 'bad quality' video to be change to VP9. The second the VP9 change took place, the video quality matched my previous videos.

I'm not sure that anyone cares but no one here seemed to have an answer for this problem so maybe it will help clear things up for future posts.

[poisondeathray]

It doesn't fully explain it. It might help explain some of the mild differences, but the test where I got you to upload the same video (the June 15 vs. 2 week later test), looked at the SAME downloaded versions (ie. h264 compression versions) locally.

[midiout]

Not sure what to tell you, maybe what you're using to download them is converting it to h.264 to save itself the trouble? Not sure.

As far as I see things, my problems are answered. The June 15th video is using VP9, the recent videos were not. I had two versions of the re uploaded video on my page, and they looked identical. Once one changed to VP9, it matched my better June 15th video, where as the other one still looks horrible and isn't using VP9. Open and shut case as far as I'm concerned.

[poisondeathray]

You're viewing with a brower and configuration set to VP9 - that doesn't mean all people get the same version when viewing at default. Some people are viewing other versions - you don't have control over that. You're not necessarily seeing the same thing other people are seeing. For example chrome almost aways gets VP9, by firefox doesn't by default. There are ways to "spoof" the browser version and get the versions you want to see, and when you use utilities like youtube-dl, you download the precise version that you want, nothing is converted

When you test these things, you need to compare "apples to apples" . So when you were looking at a different version than "someone elses" video, of course you're going to come to the wrong conclusions. Everything in your previous thread that I commented on was in a controlled situation, with the same version, downloaded directly - no conversion, and examined locally.

I'm glad you think you got it sorted out, but that one case isn't really sorted out

[poisondeathray]

Does it make sense ?

VP9 is technically better in some ways, but youtube always allocates about 10-30% less bitrate on average . Because it's "supposed" to be better, that justifies them using lower bandwidth. That nullifies the advantage greatly. If they kept the same bitrates, it would look better most of the time. If you examined a hundred different videos and compared them you would see sometimes AVC looks better in some parts of frames, some frames , but worse in others - at least that was a fact before. The key point to emphasize is that is in the way which youtube uses the codecs. When you use them locally , you can use better settings etc.. and AVC using x264 almost always looks better except at very low bitrate scenarios, and encodes about 5-10x faster. AVC also has well established HW support (GPU decoding, device compatibilty etc...)

eg. If we look at the June 15 video - The AVC version gets ~23% more bitrate . The fact that VP9 looks similar at less bitrate should indicate that it offers good compression. But that's not the whole story - it has more to do with youtube using poor AVC settings and dividing up the encoding (distributed encoding)

248 webm 1920x1080 DASH video 3433k , 30fps, video only, 85.20MiB
137 mp4 1920x1080 DASH video 4222k , 30fps, video only, 115.44MiB



I'll illustrate - these are P frame comparisons

June 15 AVC


June 15 VP9


June 29 AVC


June 29 doesn't have a VP9 version yet, as you pointed out, but notice the deterioriation on June 29 from June 15 for the AVC vs AVC version. That is what I was making a big deal about. That was the only thing that was unexplained in your other thread. Recall I said it WAS related to youtube, that it's not something in your workflow, and there were confirmed differences in the encoding settings from the two AVC versions

The reason you want to test things locally, is to see the actual video and make proper comparisons. So you're eliminating variables like browser versions, GPU acceleration /renderer etc.. You're trying to exam the actual video only, not some other issues which might be contributing or confounding . You probably don't care about all this , you just want it to look good on youtube - but what you see isn't necessarily what other people are seeing in their browsers. What version of video you get served at default depends on OS, Browser, HW configuration

AVC is much much faster for encoding than VP9. Also, if YT wanted to, they AVC versions could look much better - you can prove this by testing locally. It might sound like a consipiracy theory - but YT has strategic reasons for putting AVC in a "bad light" and to phase it out - They paid a few hundred million dollars for WEBM/VP9 to settle patent lawsuits with MPEG-LA consortium . There are no royalties to pay when using VP9 . It's almost as if they want AVC to fail, it's almost like they are "dumbing it down" on purpose