Youtube Workflow

[dellsam34]

It took some persuasion to convince the BBC Archive to switch from pillarboxed 16:9 25p to native 4:3 HD 50p: https://youtu.be/U6NnSBG_tm0?si=TBnFBgeIN7mJ03sS

I wish the BBC post some info about those transfers, such as the format and the workflow used, but they are indeed high quality transfers. I have some Betacam tapes with such contents but not sure if anybody interested in, I get very low view count for that kind of contents.

[Bwaak]

It took some persuasion to convince the BBC Archive to switch from pillarboxed 16:9 25p to native 4:3 HD 50p: https://youtu.be/U6NnSBG_tm0?si=TBnFBgeIN7mJ03sS

I wish the BBC post some info about those transfers, such as the format and the workflow used, but they are indeed high quality transfers. I have some Betacam tapes with such contents but not sure if anybody interested in, I get very low view count for that kind of contents.

Considering this is 1983, I wonder is it Betacam or 1-inch.

I would be interested to watch your Betacam tapes. ReelyInteresting has posted some.

Regarding rounded corners, they look cute, they are not too large, although I understand that someone uploading a complete frame may hate it. So far, my old Firefox version shows straight corners, Chrome shows rounded. But the Theater mode shows the complete frame, as the fullscreen. And smartphone app too.

[dellsam34]

I uploaded 2 short clips in HD and 4K from the same lossless SD capture hoping it will help some questions here about uploading in HD vs 4K for YouTube, The content may not be suited for this type of comparison, a studio recording would have been a better candidate, take a look:

https://youtu.be/p4nQde_5URY

https://youtu.be/1f90hRmVNEU

[Sunk]

Imgsli comparison. Downloaded through yt-dlp with the highest res possible.

[Sunk]

Your capture is fine from the field architecture point of view.
the histogram suffers of crushed blacks and clipped whites, probably because the card. Try to use its procamp if possible for a better distibution of the levels:


[Attachment 71271 - Click to enlarge]


[Attachment 71272 - Click to enlarge]

I know this is a slight necroquote, but I had some spare time today and so I made a new capture of that same tape. To me, the levels look much better now.

[lollo]

Imgsli comparison. Downloaded through yt-dlp with the highest res possible.

Based on your pictures (I ignore qhat you used to "equalize" them) 1080 for me; I also ignore what has been used for upscaling by dellsam34, but more artifacts were introduced at 2160

[lollo]

I know this is a slight necroquote, but I had some spare time today and so I made a new capture of that same tape. To me, the levels look much better now.

They do. But look to a sequence, not a single shot image (maybe you did it, and just post 1 image for comparison, I assume)

[Sunk]

They do. But look to a sequence, not a single shot image (maybe you didi it, and just post 1 image for comparison, I assume)

Yeah, I've gone through some of the whole footage, LGTM. I removed one of the hardware pieces (Kramer FC-400, to be precise) from my capture chain and used just my HR-S7600 + ES10. I came to the conclusion that chaining it with the ES10 blows out levels.

[Alwyn]

I took screenshots of one video i uploaded to yt in 2160p (used nnedi upscale) showing the difference between 1080p, 1440p and 2160p. It's a vhs that i tried to enchance. It's not ideal comparsion but shows bitrate difference. I dont have subscribers and only one public video. Jump from 1080p to 1440p is really visible to me.

Just looked at your imgsli; yes, I agree, the 1440P is definitely better. Look at the girl's top: the 1080 version is, relatively, a blurred mess. On almost every edge, the 1440 is crisper.

The content may not be suited for this type of comparison, a studio recording would have been a better candidate

Yes, it's hard to compare with so much action.

[Alwyn]

Rounded corners: I'm on Chrome and only the thumbnails are rounded, the main YT picture is square.

[dellsam34]

Didn't have time to work on the results, just posted the links and gone to work. Looking at the original footage I see the 4K version brought some details back that has been lost in the HD version through YT encoding, Mainly because 4K uses VP9 and HD uses h.264, but the smoothness of HD defined edges better, I used the paid version of 4K video downloader to download back the footages:
https://imgsli.com/MjAyMDU4

Comparing the samples before youtube upload made Absolutely no difference, here:
https://imgsli.com/MjAyMDUz

Here is one between 4k before and after YouTube upload:
https://imgsli.com/MjAyMDU5

I think I can safely say that VP9 is a good sharing codec via YT, but not perfect.

[Skiller]

If you add borders to a 4:3 video to make it 16:9, it will display with window-boxing (borders on all four sides) on any display that is not exactly 16:9.
Plus it's not required to stick to a certain aspect ratio on YouTube anyway, so please don't do it.

It bothers me every time when I'm watching older 4:3 music videos on YouTube and the company who uploaded it placed it in a 16:9 frame – I then get a small video in the middle of my 4:3 screen with large borders on all sides.

Therefore, this should only be done when mixing footage with true 16:9 footage.

[Krazern]

I suggest that instead of screencapping Youtube's player, you use something like yt-dlp to download a short clip from both the 1080p and the 1440p/2160p, take a screenshot in either MPC or VLC and then compare them.

So i made the comparion in the right way now. I uploaded two same short videos to yt, one 1080p and second 2160p (VHS huffyuv upscaled with NNEDI in FFV1).

I took the same frame from original video and downloaded 1080p and 2160p ones with ytdlp-interface and saved them as bmp so you can compare them fully. The 1080p video was encoded by YT in AVC while 2160p is VP9.

Here are the links to videos and frame comparsion:

https://imgsli.com/MjAzMDA5
https://imgsli.com/MjAzMDA0/0/1
https://imgsli.com/MjAzMDA4

https://youtu.be/x0PdVjCOrpc
https://youtu.be/0mMpGsrNgQI

[dellsam34]

In your case 4k vs HD difference is more pronounced compared to in my case where the difference is very minor, not sure why.

[Sunk]

I'm curious: what are the ffmpeg settings people in this thread use for their YouTube uploads? I use the following one, compressing with H.264:

ffmpeg -i in.avs -vf setsar=sar=1/1 -c:v libx264 -crf 10 -x264opts colorprim=bt709:transfer=bt709:colormatrix=bt709:force-cfr -c:a aac -b:a 192k -movflags +faststart out.mkv

Is there anything I can do to make YT go as little as possible with its own recompression? Lower the CRF, maybe?

[lollo]

For my captures, in general I upload a lossless upscaled video (HuffYUV). In rare cases I upload a compressed video with:

ffmpeg.exe -i input.avs -c:v libx264 -crf 17 -preset slow -aspect 4:3 -c:a aac -b:a 128k output.mp4

Honestly, upscaling at 1440x1080 I see little differences between lossless and h264 after YT encoding. I posted a comparison once in a digitalfaq forum thread, I do not find the link now, sorry.

[dellsam34]

Me too, lossless for now, If I decided to go 4K I may have to find a better encoding script, the files are just too big to upload raw, long uploads sometimes hang in the last step where it says processing and never finish, I had that happen to me twice where I had to re-upload the video.

[lollo]

If I decided to go 4K I may have to find a better encoding script

I think that in general "-crf 17 -preset slow" is good also if you encode a 4K lossless video prior to download. If you find something better, just let us know. Maybe some fine tuning with the options for video characteristics in x264 in accordance to the source

[dellsam34]

Is VP9 available to the public or it's just a proprietary YT codec?