Struggling with Video Quality (YouTube)

[midiout]

I've spent the last week or so trying to improve and fix my video quality. In this process I've been falling futher and further down the rabbit hole that is video encoding, realizing I have no understanding of it how it works, or in short I'm very new at dealing with this process. My source footage looks nice, my renders look decent, but after YouTube compression it becomes a blurred mess. Now I understand there's nothing you can do about the compression but what you send it to be compressed clearly has a huge impact.


Here's the quality I'm managing.
https://www.youtube.com/watch?v=29CyWSDkDAw
https://www.youtube.com/watch?v=YrptTb8AbUg
https://www.youtube.com/watch?v=cSQUK1k_uMA
https://www.youtube.com/watch?v=M81u3bxE3uU


Here's the quality I'm trying to replicate.
https://youtu.be/-LJUmnLDpDw?t=58 ( 00:58 time stamp )
https://youtu.be/-LJUmnLDpDw?t=163 ( 2:43 time stamp)
https://www.youtube.com/watch?v=HUEYJS799dI&t=41(00:41 time stamp) This clip in particular stands out to me, his 360p looks about as good as my clip in 1080p.

After all of the settings I've tired the best results have seemed to be:

• Dxtory with 'UtVideo YUV422 BT.709 VCM' 1080p 60fps outputing as AVI
• Rendered in Vegas with Windows Media Video V11, Quality VBR, 100% Quality, 1080p, 60fps, Square Pixel, de-interlacing disabled, and Re-sampling disabled.
• No after rendering compression / encoding.

Even with this being my 'best case' situtation it still looks horrible, as you can see in the example.

I've tried recording with:

• Dxtory with Lagartith, UtVideo, and Dxtory codecs. All at 1080p 60fps 100% quality AVI.
• Fraps with the ideal settings.
• Shadowplay with 50MB/s 1080p 60fps.
• Bandicam, OBS etc etc etc. I've tried every recording program I can get my hands on, with various settings.

With all of those recordings I ran them through different render settings in Vegas and Adobe Premier and uploaded a test every time I tried a new settings (bit rate, file type etc).

For Vegas I've tried:

• Main Concept AAC/AVC VBR 8MB average 14MB High and gone as far as 50-100MB. For every setting I tired it in both two pass and single pass with both CUDA and CPU rendering.
• Windows Media Video V11 100/90/85% Quality and various bitrates.
• Video For Windows AVI lossless (just to see if it would help, it didn't)
• Various video Sharpening/Saturation/Color Correction/Gama effects
• Sampling On/Off
• Deinterlacing On/Off

For Premier I Basically mimicked the settings from Vegas:

• I've stuck to H.264 because no other option seemed viable? I'm not fimiliar with many of the other settings.
• Various bit rates combinations between 8-100bMB, two pass, one pass, variable and constant, attempted them all.
• 1080p 60fps
• Sqaure Pixels
• I've basically abandoned Premier. Everything I do seems to end up much worse than Vegas once it hits YouTube. I actually prefer the workflow of Premier over Vegas so it's unfortunate that I can't seem to get it to work for me.

I'm at a total loss on where my quality is being lost. I'm thinking somehow it's in the recording process, but with this new realization of encoding I really don't know. I feel like I've done everything I can think of but have no decent results. People are playing the same games, recording and rendering with the same programs yet my videos nearly as good.

[Cauptain]

Your problem?? VEGAS/PREMIERE encode.

Export AVI lossless (UT VIDEO/LAGARITH) and encode using MEGUI/RIPBOT264/VIDCODER/HANDBRAKE/*. VARIOUS OTHERS encoders.



Claudio

[midiout]

Just to maker sure I'm understanding what you're saying. I've taken my Dxtory (UtVideo 60fps) and Shadowplay (H.264 50MBs 60fps) footage and put it into Vegas using these project settings. Rendered it in "Videos for Windows" using Lagarith, with these settings. Then put it through Handbreak with these settings.

I'm waiting for the upload now to see the results. It seems to be taking ages to process when I use Handbreak, which is fine, but it makes me wonder if there's better advanced settings I could be using.

[Cauptain]

Just to maker sure I'm understanding what you're saying. I've taken my Dxtory (UtVideo 60fps) and Shadowplay (H.264 50MBs 60fps) footage and put it into Vegas using these project settings. Rendered it in "Videos for Windows" using Lagarith, with these settings. Then put it through Handbreak with these settings.

I'm waiting for the upload now to see the results. It seems to be taking ages to process when I use Handbreak, which is fine, but it makes me wonder if there's better advanced settings I could be using.

Yes or just download DEBUG FRAMESERVER for VEGAS, install and export directly to any encoder (I like RIPBOT264).

All my videos are so.

Look: https://www.youtube.com/watch?v=pEn4o-6IzEA



Claudio

[midiout]

The results are a little better with this process, but still kind of iffy and inconsistent.

For example there's this:
https://www.youtube.com/watch?v=OiPzc0YcprA

The first two clips look pretty good with the motion, not to much of a problem. When I saw the first two clips I was pretty excited, thinking I've come to a resolution. So I was pretty satisfied with the first two clips so I tried it again with another clip. Sort of a worst case scenario to see how it would work. Bright scene, high motion.
https://www.youtube.com/watch?v=-T_iz23jWjw

Here it falls back to my older examples. It seems like a brighter scene with motion just ruins the video every time. It usually looks decent in dark areas but once I'm in daylight it all starts to blur together. The usual give away is the names/text on the top left hand side. For all of these clips I followed the process from my previous post (lossless to handbreak).

The strangest thing about all of this is that last week I didn't have any of these problems. I was recording with Dxtory (Lagarith 30fps) and rendering from Vegas with Windows Media Video V11, Quality VBR, 100% Quality.

This was giving me ideal results, here for example:
https://youtu.be/17IyANQp0AU?t=49

Then a few days later; using the same process, the video's just turned to crap once I upload them.


Edit: I guess I didn't let it process long enough, even though it was letting me go 1080p60. It's looking pretty much as good as I can expect it to after YouTube compression. Thanks for the advice Cauptain (Claudio), I appreciate it.

[midiout]

Turns out it actually hasn't changed anything. My videos are still noticeably lower quality than the examples I've provided. I'm really not sure what to do?

[Cauptain]

Youtube recompress any video uploaded...Ever recompress. Its the problem.

Your method is OK.

Try higher bitrates and make a test.




Claudio

[midiout]

I've tried higher bitrates, I've even recorded small clips with Dxtory and uploaded those massive files directly. I understand that YouTube compresses the video no matter what, but the video's I linked at the top were also compressed in the same way but look fantastic.

[newpball]

All we need to get is a few seconds upload of the source and a few seconds of what you eventually send to YouTube.

[midiout]

When you ask for uploaded source I assume you mean not on youtube? I'm not sure where else to upload it for it to be of any use.

Well this is the source uploaded to youtube. It looks horrible once uploaded, but basically looks like I'm playing the game when played directly from my computer. Shadowplay 60fps 1080p 50mb/s

This is that same footage run through Vegas as if it was part of an edit: https://www.youtube.com/watch?v=zrRV4_Z-1S4&feature=youtu.be
Project Settings
Render Settings (Video for Windows .avi)
Handbreak settings

I'm sure all of these other details after the source video were pointless, seeing as the source video looks to be the problem itself?

Talking to this person he's told me he achieved that quality by recording with Shadowplay (like I have), compressing it by 'about half' with handbreak and uploading it. His footage looks perfect.

[newpball]

Upload them for few seconds on this website.

I want to see the source and the cooked version to see if the culprit is actually YouTube or if it is an overly compressed or otherwise butchered video.

[themaster1]

Try cuda h264 (now defunct) or nvenc h264 with mediacoder, with not many ref/b frames and a high bitrate 12k or upper in a .mkv container and pcm sound. I always had good results on youtube with such method.

[poisondeathray]

It doesn't much what you do, YT always re-encodes

The reason the other ones might look slightly better is the content is different. That's the bottom line

That's how compression works. When you have lots of action, details, motion - it's more difficult to compress. Compression works by storing the differences between frames. If you have big differences (e.g. lots of motion), that means it needs lots of bitrate to preserve it. And at the bitrates YT uses, it will "fall apart ." because there isn't enough. Try uploading a scene where you play a sniper with very little motion - you will notice it look better because the compression isn't "taxed"

That's why the one you liked in the dark scene looked better. It's mostly because it was darker there are fewer "shades" to represent, and less action. If you have stretched color pallete with lots of colors, brights, darks - it's more difficult to compress

If those users uploaded your video, it would look the same as you uploading it . They are not getting preferential treatment, they aren't doing anything special

Forget about trying different encoders - You can upload a lossless version and it will still look about the same. You have no control over how youtube deals with it, and YT's re-encoding is the limiting factor

You can try to optimize it , just as blurring, denoising - those types of things help compression on YT's end - but the benefit is very small in the end

You can do things like editing: interspersing sections with little action, breaks, commentary - that does help a little, but again very small in the end

If you want it to look good, use another hosting site

[midiout]

I understand that Youtube compresses and encodes no mater what I do. But the examples I tried to provide are pretty similar situations in terms of lighting and movement.

My example:
https://www.youtube.com/watch?v=YrptTb8AbUg

What I'm trying to achieve (the time stamp is important 2m42s):
https://www.youtube.com/watch?v=-LJUmnLDpDw&feature=youtu.be&t=163

It's the same game, at the same location, of the same map. They both have similar amounts of motion and action, yet one is obviously MUCH clearer than the other.

Or for example, before I ran into these problems.

This is a video I uploaded two weeks ago. It looks more or less exactly how I want it to.
https://www.youtube.com/watch?v=17IyANQp0AU

If I now render and upload the same video, from the same project file, with the same source files and the same render settings. It looks like this:
https://www.youtube.com/watch?v=FGxSYtA5dt8

Understandably it's not THAT much worse, but it's still worse... for no real reason. That example was even sort of bias because that clip looks much better than most of my current uploads. Where as my previous uploads were just fine under the same circumstances. I thought that maybe it's something to do with the full video vs a short clip, but after uploading the entire video again there was still no improvement. This is all a new problem I'm having. Last week my videos looked exactly how I wanted them to every time I uploaded one. But now it's hit or miss, 80% of the time they look horrible, with the exact same setup as my videos from last week.

Edit: Here's the link to a source video as requested.
http://files.videohelp.com/u/243643/Dxtory%20UtVideo%2060fps%201080p.avi

[poisondeathray]

It does have to do with long vs. short. Are you saying you uploaded the exact same clip and it still looked worse ? Please provide the links

Did you examine parts of the clip before you linked ? There are periods of inactivity. Ducking behind containers etc.. This is what I mean by editing adding breaks etc... Of course you can't do that in "normal" gameplay. If you have continuous action, it never has a chance to recover. Continuous actiion without breaks will always look bad out youtube. Even adding something like an intro, outtro will improve the quality, because there is a break at the begining and end. This was demonstrated in previous posts

[_Al_]

How about to try to resize it to 960x540 and then upscale it to 1080p with simple point resize. Then upload to YouTube.

I doubt that those 1080p game videos have real 1080 resolution anyway. So if "audience" wants 1080 resolution as a gimmick I'd try to do that, also YouTube might give it more bitrate than it would distribute for 540p. So this way you'd have blocks, but very little macroblocks, not broken picture etc.

[midiout]

Sorry for the delay, I must have deleted the video so I had to re-upload it. Both videos were recorded with Dxtory using Lagarith at 1080p 30FPS. Rendered with Sony Vegas with Windows Media Video V11, Quality VBR, 86% Quality and no further compression other than you tube.

Here's the before:
https://www.youtube.com/watch?v=17IyANQp0AU
https://forum.videohelp.com/images/imgfiles/RbB4ix4.jpg

And the After:
https://www.youtube.com/watch?v=eVTsneG4vbw&feature=youtu.be
https://forum.videohelp.com/images/imgfiles/2dmmpJE.jpg

Again this is the same project file, project settings, source files and render settings. My videos would consistently look like the before clip every time I uploaded them, no matter the content or length. Now they consistently look like the second video, no matter the content or length. You could see my enemy's eyes in the first example, now you cant even tell if he has a face.

[poisondeathray]

You messed something up because the durations are different . Your overlays are slightly different, different text, but that shouldn't affect it too much. But you're screenshot frames aren't even the same. You could be comparing an I frame vs. B frame for example - that' s just for future reference when making proper comparisons



BUT - when examining the downloaded videos, the newer version is definitely worse, a lot worse. Almost everywhere (there are a few frames where the new version is slightly better, but very few). And the "old" version isn't that "old", June 15,2015

So unless you did mess something else up other than the duration, something different is going on between June 15 and now...

Sometimes it takes a while for YT to process a higher quality version, but I don't think thats the reason here. Maybe this is happening to everyone, everywhere, across the board for YT ?

[_Al_]

try this: https://youtu.be/qBnwCBoxlbQ
I encoded your sample and uploaded on youtube

action is much better but describtions, small fonts are obviously bad

[midiout]

The only thing causing them to be different length is that I render loop region in Vegas and I manually set the region before re-rendering and uploading. As far as the frame differences for the images, it was just for a quicker example to show my problem. It wasn't the exact frame because I was struggling to pause the video at the exact same times in youtube. I feel like that doesn't really matter though since it gets the point that I'm trying to make across. The quality across both videos is essentially summed up in those two screen shots even if they're not at the exact same frame.

[poisondeathray]

Understood, but the text overlays are different, duration is different - so no it wasn't the exact same project.

Nothing against you personally, but when people make a bunch of little mistakes, it make me wonder if there might have been a few others. Maybe you didn't actually use the same render settings, maybe slightly different processing like one was deinterlaced, etc...

If you want to get to the bottom of things, you need to do more proper scientific testing, where all the variables are controlled.



But assuming you did everything else correctly, those are significant quality differences (I examined the downloaded versions, not from YT viewer) that cannot be explained by those things alone. It makes me wonder if some big change has happened since June 15 at YT .

[midiout]

Well no one else's videos seem to be having this issue. I don't know what you mean about my overlays being different, other than a the fonts that pop up, and that's simply because I've removed that font from my computer (today; after these problems existed). Everything else is exactly the same, I opened the same project that was last saved before I rendered it for upload previously. The interlace, resampling, and render settings are all the same. Everything about the project other than it being a few seconds shorter; based on me quickly selecting a star and end point before re-rendering it, is the same.

So assuming I'm accurate when saying all of that, and the idea that no one else seems to be experiencing this problem, am I basically being told there's no options for me to resolve this?

[poisondeathray]

I don't know how you can imply that it "only affects you"

Take some random exact same video that you upload on your account, and have someone else upload it as well. If you get something lower quality and different, then yes it only affects you. Until you do that test, it doesn't "only affect you."

But there is something definitely funky going on with the June 15 vs. Today test. That difference is just way to large for a bunch of little differences to explain

[midiout]

I'm not saying it only affects me. I'm trying to say something is happening on my end,with my computer or process, that is now ruining my videos; when it previously wasn't a problem, and that it does not seem to be YouTube related. I'd also think it's not related to the recording process because the same source files I previously used also turn out horribly. I'm thinking something Vegas related is causing this problem but I'm not knowledge enough on the program or the rendering/encoding process to determine what that problem is.

[poisondeathray]

I'm not saying it only affects me. I'm trying to say something is happening on my end,with my computer or process, that is now ruining my videos; when it previously wasn't a problem, and that it does not seem to be YouTube related. I'd also think it's not related to the recording process because the same source files I previously used also turn out horribly. I'm thinking something Vegas related is causing this problem but I'm not knowledge enough on the program or the rendering/encoding process to determine what that problem is.

How do you know it's your computer or process? How do you know it's not youtube related ?The "best" you can do is upload a decent quality video. That's what everyone else does. It doesn't make much of a difference what you do as long as the video quality is ok. Youtube doesn't care what your computer or process is. All it "sees" is the video you upload. It decodes that. So take that and have someone else upload it . If you both get the same thing, you can probably assume it affects everybody - it's probably a global change

It IS youtube related, at least partially. I'm still looking into it, but for example, the June 15 version had 3 reference frames, the Today version had 2. (IN general, if you have more references frames, it will yield higher compression efficiency , so higher qualty) . So yes, there are youtube encoding differences. If we assume that the different duration and text and possibly something else don't make a difference, then yes it is youtube related

Or was it something that triggered the more reference frames (and possibly other differences I haven't looked closely into yet) ? We don't know because you didn't upload the exact same video.

That is why you should be doing controlled testing. If you upload the exact same video (not different length, not slightly different), and there are still differences, then you know for sure that it's only youtube .

[midiout]

Well I could record a new video, render it and upload it twice to see but that really wouldn't make much sense the way I see it. I'll be uploading two videos which will have presumably poor quality. This is the process I've been going through for the last week. Where as my 'control' in a way would be my old video. It didn't have these problems with that video or anything else I uploaded. I've been trying to get back to the ground floor by trying various setups and settings in order to get on par with that video again.

As far as it being my process or computer and not youtube. Assuming for an extreme example my hard drive or ram were screwed up in some way, could the render may end up being a mess; maybe not visually during playback, but in the back end? If that was possible, it would lead to a poor upload for me or anyone else I ask to upload the video for me.

Or there's the idea that other youtubers small and large, seem to be having no problems uploading footage of this same game. In which case I'd be looking to ditch my process entirely and look to find something new and improved on what I'm doing (my main reason for reason for creating this thread). Sure you can say their videos have more down time, or still frames, different frequency of dark/light/moving scenes. But generally the product we're putting out is the same and they have a consistency in their video quality no matter the amount of motion or colors. Their videos look great if the entire thing is nonstop action and motion, in dark areas, in light area. They put out a good consistent quality and I'm looking to achieve that. Though I do understand my videos having different content causing differences in quality, I also have no idea what their recording and rendering settings are.

So opposed to the idea of trying to determine what is similar or different from my two uploads, maybe it's in my best interest to ditch my previous attempts and start over. Maybe I should try what would theoretically yield results up to the standard I'm looking for (though I don't know what that process would be). I do appreciate anyone who's attempting to assist me , I feel like I may be coming across aggressive and I don't mean to be. I'm just trying to get on a level way of thinking about this problem/solution, though I don't have the knowledge about video creation/encoding that anyone else here does. This entire situation has been sort of annoying, going through this trial and error for the last week I feel a bit defeated.

[poisondeathray]

Well I could record a new video, render it and upload it twice to see but that really wouldn't make much sense the way I see it. I'll be uploading two videos which will have presumably poor quality. This is the process I've been going through for the last week. Where as my 'control' in a way would be my old video. It didn't have these problems with that video and I've been trying to get back to the ground floor by trying various setups and settings in order to get on par with that video again.

That's not what I'm asking.

What I see is a significant drop in quality if we assume what you uploaded was similar to June 15. This is abnormal for sure.

If you uploaded something similiar in quality today to as what you did on June 15, then you should get similar results. You don't . So that implies something has changed on youtube's end, not what you did or your computer or your process - does that make sense ?

Getting someone else to upload a video, is to circumvent youtube's detection. You can't upload the same video to the same account (it will detect the duplicate). Additionally, there have been some theories that users in certain countries get better service/ better encodes than others. Maybe there are more servers in "xyz" location etc.. (having more servers, more distribution for a single encode will reduce the quality) . None of that has been conclusively proven

As far as it being my process or computer and not youtube. Assuming for an extreme example my hard drive or ram were screwed up in some way, the render may end up being a mess, maybe not visually during playback, but in the back end. This would lead to a poor upload for me or anyone else I ask to upload the video for me.

No , if the render is a mess, it would be a mess for everybody visually during playback. It means the video is a mess.

But you're not uploading a mess. You said it looks fine locally. You're supposedly uploading something similar to what you uploaded June 15. If youtube was the same, then you would get the same (or very similar) results

The problem isn't what you are doing. As long as you have something that looks decent locally before you upload.

Or there's the idea that other youtubers small and large, seem to be having no problems uploading footage of this same game. In which case I'd be looking to ditch my process entirely and look to find something new and improved on what I'm doing (my main reason for reason for creating this thread). Sure you can say their videos have more down time, or still frames, different frequency of dark/light/moving scenes. But generally the product we're putting out is the same and they have a consistency in their video quality no matter the amount of motion or colors. Their videos look great if the entire thing is nonstop action and motion, in dark areas, in light area. They put out a good consistent quality and I'm looking to achieve that. Though I do understand my videos having different content causing differences in quality, I also have no idea what their recording and rendering settings are.

It doesn't matter. That's the least important difference. You seem fixated on what your process or their process is as if it will make a big difference. It doesn't.

There are a bunch of things you can do to optimize, reduce the compression requirements for YT to handle, but the end result is minor.

So opposed to the idea of trying to determine what is similar or different from my two uploads, maybe it's in my best interest to ditch my previous attempts and start over. Maybe I should try what would theoretically yield results up to the standard I'm looking for (though I don't know what that process would be). I do appreciate anyone who's attempting to assist me , I feel like I may be coming across aggressive and I don't mean to be. I'm just trying to get on a level way of thinking about this problem/solution, though I don't have the knowledge about video creation/encoding that anyone else here does. This entire situation has been sort of annoying, going through this trial and error for the last week I feel a bit defeated.

To be blunt, you're trying a bunch of things and wasting your time. It's not how compression or youtube works. (and you have tried a bunch of things, but nothing worked , right ? )

My guess is that there has been some change in YT recently. At first I wanted to brush everything off to content differences for those other videos (and it does make a difference, it's easy to prove and has been demonstrated before) but that difference in your June 15 vs. Today test is too large, assuming the "today" test was done correctly - it suggests something else is going on

Perhaps they are distributing their encodes more (cutting up into more pieces across more servers), that will reduce quality. But there are fewer reference frames in Today's encode, and that does reduce quality. Why did it do that today, but not June 15 ?

[midiout]

Well since I no longer have the exact same rendered videos from my previous uploads I have no real way of saying it's the absolute EXACT same footage. But it is the same source footage rendered with the same settings.

About the render being a clean(locally) and getting to youtube and being inconsistent. I feel like there has to be more to it than that. For example, I have a video that looks quite crisp there's a very slight noise to it, based on me using a different program to record it. With that noise the gamma is very slightly brighter. Now in terms if saying it looks fine locally, it's very clear, no distortion or artifacts, everything is very defined but once that slight noise hits you tube it's a total mess. Maybe my idea of what looks good during playback isn't in par with someone who's more familiar and knowledgeable with video footage. In a similar way that an audio engineer would know how to EQ and mix a song where as a consumer already thinks it sounds great because it has a lot of bass.

I almost want to drop the idea of Youtube having changed something and try to move past that. If they have changed something it doesn't seem to be affecting anyone else. Other YouTubers uploading similar content of the same game after June 15th with fantastic quality. I feel like they didn't just get 'lucky' since all of their videos look great, every time. In that case I'm basically asking, 'what do I need to do in order to achieve the same quality of this person', or 'what can we assume these people have done in order to achieve this?'. A lot of people are out there making videos now not having the problem I amm so I want to theorize what they are doing differently than me in order to get these consistently great videos (even after June 15th).

[newpball]

We are never ever going beyond the "it may me all YouTube's fault, or it may not" if you do not provide the captured source and the cooked version you upload to YouTube. 10 seconds of each will do!

My guess is that the video is already butchered when it is uploaded to YouTube, the fact that the OP apparently refuses to provide samples only affirms it.

[poisondeathray]

Definitely, noise impairs compression efficiency. Recall that's how video compression works, by storing the differences between frames. Noise by definition is random, and will make adjacent frames appear to be "different" in compression terms

But if we extend this to youtube - it has to be quite noisy to make a big difference. If you want someone to check it out, upload something that you would have uploaded to youtube

But that is one of things that people do to optimize mentioned eariler - they denoise to improve comrpession efficiency, they add things like motion blur. A video that is more blurry is easier to compress for YT than one that is razor sharp. The "lower" quality video will often end up looking very slightly better . Again those are minor differences, not major.



But 99.9999% of youtube user just upload it as is. They throw it into something like handbrake.

The quality differences between the other videos can all be explained by content differences. For example, pauses, peripheral overlays (in the 2nd video), less action, dark video with limited colors (the 3rd video) . This is all easy to prove, and has been proven before. It's how compression works.

However, the June 15 vs Today test cannot be easily explained. Those are unexpected results, assuming you uploaded videos of similar quality .