Should We Upload To Youtube Precompressed Or Allow Youtube To Compress Video For Us?

[Superballs]

I've noticed that this question has been asked a lot.

I've had a recent video that I just finished and decided that I would upload it twice, with two different bitrates, and we can try to objectively decide which is the better way.

I have not had a chance to view and compare my video quality on these yet but will chime in my "post view" opinion later in the comments. Currently (just to present my hypothesis), I am of the opinion that pro grade editing software is going to do a better job at encoding the video at lower bitrates than YouTube will, and if we upload our videos at a high bitrate, YouTube's re-encoding will mangle them.

So, for all of our consideration, here is my video on Road Rage.

Here is the publicly uploaded video that was encoded at approx 12.5mbit (target) max bitrate was 14 and it was encoded VBR, 2-pass. Mods, if possible to make it easier to get to, if you could edit my links to proper URLs, but it seems that the way to get a link pre 20-post is to just include the end.

watch?v=7rdxCwj3Pjk

This URL is the exact save video, from the exact same project with no changes, and the encoder's output settings was CBR at 28mbit/s, which very closely matches my camera's output at 1080p @ 30fps. This is an unlisted URL, so cannot be searched and is only up for testing purposes.

watch?v=5va2iNFcXPo

As I've said, I won't be able to chime in as to my opinion of which is better quality until later, I'm hoping that some other discerning eyes will be able to take a look and decide.

The one titled "High Bit Rate" had a file size of 2.77GB and the one that was publicly published had a file size of 1.22GB

I will also, later tonight, try and do an in-depth analysis of areas where my local files were radically different, and will post screenshots where possible at specific times. After that, I will try and find the closest frame possible in the YouTube copy to compare to, to try and see what differences are in the YouTube copy, vs the local files, for each one and see which one is more drastically altered from the original.

 

[Superballs]

So, I've had some time to go over the videos and found one spot in both where the compression artifacts were apparent.

I'm going to upload the pics, and they are very similar. First one will be the precompressed video and the second will be the high bitrate video

I captured the frame by playing the video in one version at quarter speed and pausing as close as I could catch to the 3:07 mark (a couple frames past it)

I grabbed a screenshot using FastStone Capture, capture windowed area and I captured a section from a specific, identifiable pixel, at the top of the picture (very top of a telephone pole). Capture size was 857 pixels by 338 pixels (bottom center of a car tire).

I then went to the high bitrate video and played at quarter speed. Kept trying frame by frame until I saw the frame that most closely matched and tried to select the exact same pixel at the top of the telephone pole, and ensured that my capture was 857x338

I resized both images to 200% using the FastStone editor.

I've personally noticed that while the compression artifacts are certainly different, it's really hard to say that one is necessarily "better" than the other.

Watching through the video, I still feel that precompressing is a better approach, as while the overall quality will be a bit lower, (IE the best quality frames will be lower quality but the lowest quality frames will be of a higher quality), the resulting effect is that the compression artifacts end up a lot less jarring and noticable due to a smaller change between the best and worst quality.

The higher bitrate video's best looking frames were fantastic, but the worst looking frames were horrid and the effect is that my eyes could really pick out the artifacts better.

The potential quality in the high bitrate video was astounding but when it gets bad, it gets worse. The higher "contrast" between good an bad makes me feel that it's the worse way to go.

That said, I think this is going to be a very very subjective thing, and may even differ based on the content.

Anyway, the images are below, I would suggest saving them both to a folder, so you can use the Windows Photo Viewer (or your mac/linux equivalent) to switch between the two (they are very very close) and see what you think.

 

[Superballs]

Just to show that YouTube is, in fact, the culprit here, I've uploaded (close to) the same frame and size (was less careful really) showing the encoding out of Premiere, both at 28mbit/s and at 12.5 or so (CBR and 2-pass VBR respectively).

I've also attached a screen of the raw footage as the third image.

 

[scooterwuf]

Visually what I'm seeing is a reduction in Blacks from the top down, or should I say it's being added if the bottom image is your from your raw footage (not tone confused with RAW footage), and the top two have been processed/compressed.

- Wuf

 

[Superballs]

Visually what I'm seeing is a reduction in Blacks from the top down, or should I say it's being added if the bottom image is your from your raw footage (not tone confused with RAW footage), and the top two have been processed/compressed.

- Wuf

Bottom is the footage straight out of the camera.

It might be hard to judge the entire frame though, it's really just a small portion.

If you check out the video, you can probably pick out where the grab is from, at the 3:07 mark.

Also, and not sure if it's going to make a difference tonally, but the footage right out of the camera is 60fps, and the exported footage is 30fps (I messed up my sequence settings )

 

[Dark Rider]

I did the same test as you a while ago and I settled on using CBR via adobe media encoder as the difference was barely noticeable for me between VBR2 pass and CBR. VBR2 pass took well over 16 hours to render on my pc.

CBR looks just as good but gives a smaller file size. I use +/- 10mbit depending on the video length. I normally aim for a 1 gig file max.

 

[Superballs]

I did the same test as you a while ago and I settled on using CBR via adobe media encoder as the difference was barely noticeable for me between VBR2 pass and CBR. VBR2 pass took well over 16 hours to render on my pc.

CBR looks just as good but gives a smaller file size. I use +/- 10mbit depending on the video length. I normally aim for a 1 gig file max.

I agree there I think. I'm going to shoot for around 10mbit, just to try and make sure I'm well under the YouTube limit. I'm hoping to avoid any re-encoding at all.

 

[WingManGT]

I export in CBR 20mb and I'm happy with the results and doesn't take ages to render.

 

[Superballs]

On my next upload, I'll try 20mbps and 10mbps.

I'll compare the quality both before and after the upload, and I'll try to be as frame accurate as I was (or more so if I can.

I know that YouTube supports mp4, so re-encoding shouldn't be a thing, unless they don't like the h.264 codec and re-encode that, but I've used the YouTube downloader (was the only way to test YouTube 4K videos on my 4K TV which was manufactured during the advent of the h.265 HEVC vs VP9 format war and doesn't support VP9), and I'm pretty sure that they were all h.264 inside an MP4 container.

 

[AmpdMoto]

Youtube only allows a certain mbps for each video frame size so anything you choose beyond that is getting compressed regardless. 4k gets a higher bitrate than 1080 but makes for large files so it's really up to you.

Just google youtube bitrate settings and you'll see what the max they give to each resolution is so you can pick what works best for your settings.

 

[Superballs]

Youtube only allows a certain mbps for each video frame size so anything you choose beyond that is getting compressed regardless. 4k gets a higher bitrate than 1080 but makes for large files so it's really up to you.

Just google youtube bitrate settings and you'll see what the max they give to each resolution is so you can pick what works best for your settings.

Yeah.

1080p is 12mbit/s

4K is a nice workaround though, you can upscale a 1080 video to 4k and it will look great coming from youtube, assuming your machine can run it