Author Topic: eevBLAB #16 - 4K Render Youtube Experiment  (Read 20781 times)

0 Members and 1 Guest are viewing this topic.

Offline SteveLy

  • Regular Contributor
  • *
  • Posts: 220
  • Country: au
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #25 on: October 09, 2015, 02:12:31 am »
I downloaded the vids and had a look at identical frames from the 1080p-youtube/4k-upload and 1080p-youtube/1080p-upload (EEVBlog2) versions. The compression noise is certainly different but there is not much in it in terms of quality; some bits look perhaps a whisker better on one, others on another.

But this video is not a good one to test the idea on. It's hard to tell what's more noise vs more detail and processing blur from genuinely more accurately represented smooth textures in the background.

If the upscaled-4K upload does improve 1080p viewing quality then it could be reserved for special videos that warrant that level of detail. There is no need for it in normal talking head, DaveCad, teardown vids. Good camera focus, not shaking the camera and what's in front of it are a lot more important for pleasant viewing experience. And I hate the idea of wasting all that extra processing power and bandwidth just to get a bee's dick of an increase in quality.

In my limited youtube experiece one could maximise quality for a given resolution by compressing the video in a format and with settings that youtube was happy to accept as is without reprocessing. (But I haven't tried that in a while; maybe it does not work any more.)

FWIW, here is a frame from both 1080p videos (click thumbnails for full 1080p frame PNGs):

Dave in 1080p from 1080p:


Dave in 1080p from 4K:

« Last Edit: October 09, 2015, 03:36:13 am by SteveLy »
 

Online coppice

  • Super Contributor
  • ***
  • Posts: 3758
  • Country: gb
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #26 on: October 09, 2015, 02:16:30 am »
Well that's just showing off. Lol

I wonder if the You Tube player varies though... or maybe the content isn't at all resolutions in all countries.

I'm in the UK.

Sent from my XT1039 using Tapatalk
Youtube does serve us mostly from local servers, and we have 1Gbps internet so we have no bandwidth restrictions to those local servers. I'm in HK.
 

Offline The Magic Rabbit

  • Regular Contributor
  • *
  • Posts: 108
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #27 on: October 09, 2015, 02:18:30 am »
My connection is 24Mbps... Which in the UK is considered fast.

Sent from my XT1039 using Tapatalk

 

Offline GNU_Ninja

  • Regular Contributor
  • *
  • Posts: 200
  • Country: gb
  • Mostly Harmless
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #28 on: October 09, 2015, 02:20:09 am »
Just tried to watch that in 2160p 4K ... Crappy works WiFi and my Acer Chromebook 11 ... The Chromebook worked fine but the painfully low bandwidth WiFi link fell over and choked on the data   :(

There wasn't an option to watch at 1080p 4K in my browsers YouTube settings, so I selected 2160p 4K  :-//

Personally, I usually watch in 720p HD (or lower) on my laptop. Internet 'speed' in the UK is piss poor  :)
« Last Edit: October 09, 2015, 07:07:02 pm by GNU_Ninja »
 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3092
  • Country: us
  • L.D.A.
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #29 on: October 09, 2015, 04:38:59 am »
My connection is 24Mbps... Which in the UK is considered fast.
I have 3Mbps (poor man in Calif) .  I am glad of this test because a while ago I stopped watching Dave's videos because of my bandwidth and now I see YT resizes them to 480p (Auto setting) so I can watch them now :).  At 4k my computer just locks up  |O.  I can force it to 720p and it plays, but I will let it use Auto unless I need to increase the res.  Then if I really really need high res, I download the video in high res and watch it with VLC.

Dave if you see this:  why do you use Handbrake, can't you take the file from your camera and load it directly into Vegas?

YouTube and Website Electronic Resources ------>  http://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline The Magic Rabbit

  • Regular Contributor
  • *
  • Posts: 108
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #30 on: October 09, 2015, 04:41:49 am »
I think only the latest version of Vegas supports 4K. The Canon HG series will import into Vegas 11 at 1080p.

Sent from my XT1039 using Tapatalk

 

Offline ez24

  • Super Contributor
  • ***
  • Posts: 3092
  • Country: us
  • L.D.A.
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #31 on: October 09, 2015, 04:56:07 am »
I think only the latest version of Vegas supports 4K. The Canon HG series will import into Vegas 11 at 1080p.

Sent from my XT1039 using Tapatalk
thanks
YouTube and Website Electronic Resources ------>  http://www.eevblog.com/forum/other-blog-specific/a/msg1341166/#msg1341166
 

Offline hayatepilot

  • Frequent Contributor
  • **
  • Posts: 257
  • Country: ch
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #32 on: October 09, 2015, 05:19:32 am »
I think the High Frame Rate looks much, much better than the 4K upscaled version.
Please stick with 50p, it's not worth the extra time it takes for rendering and uploading.

Keep the videos coming!  :-+

Greetings
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 4386
  • Country: gb
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #33 on: October 09, 2015, 07:49:59 am »
Two lossless .bmps, screenshots from each of the two videos, on a 4k monitor. I can't see the difference from a purely still picture perspective.

24MB each when you select the download link or view the original.

1080p: https://onedrive.live.com/redir?resid=2EDA80327053C74D!49896&authkey=!APFzt8UC-2XHWmM&v=3&ithint=photo%2cbmp

2160p: https://onedrive.live.com/redir?resid=2EDA80327053C74D!49897&authkey=!AI5rPOWgd2Dxrlo&v=3&ithint=photo%2cbmp

Edit: attached a 4x zoomed in photo, 2160 on left, 1080 on right, looks like more artefacts on the 1080 to me. In practice, I couldn't tell, and I had to try hard to find an example where I could identify a difference.
« Last Edit: October 09, 2015, 07:59:01 am by Howardlong »
 

Offline tbrucenyc

  • Newbie
  • Posts: 1
  • Country: us
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #34 on: October 09, 2015, 08:29:46 am »
How would anyone "see" the 50 fps rate unless there was a lot of action?
I could not see the difference between the two 1920 x 1080@30
Both looked great!
You don't have a lot of screen action so you don't need a high frame rate.
Besides:
My microprocessor dropped frames at 50 fps so I've been having to watch at 720...
Whaa Whaa Whaa Whaa
(Fiber Optic Internet 50Mbps  1920 x 1200 x 59 monitor)
Image is of Right click; Stats for nerds... Connection speed says 45171 Kbps
« Last Edit: October 09, 2015, 11:04:33 pm by tbrucenyc »
 

Offline plexus

  • Contributor
  • Posts: 35
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #35 on: October 09, 2015, 11:30:46 am »
I am watching via a 15Mbps DSL connection to an Apple TV via HDMI into a 50" HP Plasma TV (no it doesn't do RPN). I carefully went back and forth comparing and the upscaled 4K looks much better. its crisper, seems to have a wider colour gamut. in the 1080p version there are highlights on your face that blow out where-as on the 4K there is some detail there and they don't blow out. I can try and take some pics but it will be a challenge with the moire the TV mask will resolve.
 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 720
  • Country: se
  • Hobbyist
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #36 on: October 09, 2015, 12:09:10 pm »
I only have a 1920x1080 monitor. Both 2160 and 1080 looks the same to me, although in the posted still images in this thread there does indeed seem to be a small difference. 2160p stutters even though my computer and broadband should be fast enough (in theory), not sure why this is yet but it's not watchable as it is.

I think there is a big improvement with video shot at 50 fps over 30 fps though.

So personally I would much prefer 1080@50 over 2160@30.
 

Offline rs20

  • Super Contributor
  • ***
  • Posts: 2060
  • Country: au
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #37 on: October 09, 2015, 01:58:01 pm »
I'm still not 100% convinced that these still image comparisons are valid, I mean, are we sure that we aren't grabbing an I frame from one video and a B frame from the other? But based on other anecdotal evidence, we have to conclude that the 2160p30 viewed at 1080p30 is marginally better-looking than the 1080p30 native. I, for one, think 1080p50 looks far better than the indistuingishable (to me) 1080p30 and 2160p30. Even when it's just a talking head shot, it's just so fluid and life-like, I really like the 50fps.

As for how a smaller file (the 2160p transcoded to 1080p file) can appear better quality, here's some brand new speculation: If you allow a video compressor more time do its compressing, it can achieve a higher quality for a given fixed (or even slightly smaller) bitrate. Two-pass encoding is just one example of this, but there are presumably many parameters to fiddle with. Given that only a negligible fraction of files uploaded are 4K at the moment; YouTube can afford to spend a lot more CPU time transcoding each video. By contrast, every % of performance they can squeeze out of the 1080p compressor is worth gazillions of dollars, so that's going to be optimized down to a cost. I think that could be one explaination for what we're seeing here.

I only have a 1920x1080 monitor. Both 2160 and 1080 looks the same to me, although in the posted still images in this thread there does indeed seem to be a small difference. 2160p stutters even though my computer and broadband should be fast enough (in theory), not sure why this is yet but it's not watchable as it is.

Wait, are you playing the 2160p video AT 2160p? Dave's OP specifically instructs us to play the 2160p video at 1080p:

Can you see any difference between this video at 1080p and this other one uploaded at 1080p?

If so, that's fascinating that the smaller file would be more demanding to decode AND higher quality to watch.
 

Offline SL4P

  • Super Contributor
  • ***
  • Posts: 2034
  • Country: au
  • There's more value if you figure it out yourself!
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #38 on: October 09, 2015, 02:58:51 pm »
In my old fashioned opinion, if the content justified anything better than 1080 - I'd be happy to use it, but anything short of special events with well-defined & lit shots with broadcast lenses and ... seems like overkill.   Even 1080 is a bit of a stretch for a close-up shot inside a piece of 20-year old gear, or a whiteboard lecture.

Everything is nice, but the supporting overheads (set, lighting, make-up, encopding, conversion, bandwidth) - seem to make it largely irrelevant - added to the fact that most of the punters wil be watching on ADSL at 10Mbps or less.

For that 10-20% of casual viewers capable of seeing the 4K (rendered properly) in real-time, I'd rather you spend your time with the family or doing more benchtop videos!
Don't ask a question if you aren't willing to listen to the answer.
 

Online NiHaoMike

  • Super Contributor
  • ***
  • Posts: 4518
  • Country: us
  • "Don't turn it on - Take it apart!"
    • Facebook Page
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #39 on: October 09, 2015, 03:50:46 pm »
I do see some improvement with 4K upscaled at your end as opposed to 1080p upscaled to 4K with my GPU, but it's not by a whole lot. I'm sure it's the bitrate that made the actual difference. With high bitrate 1080p such as broadcast TV, the upscaling works very well indeed.

For those with just a 1080p monitor, what about try comparing 4K downscaled to 1080p and plain 1080p? Any GPU of the 650 series (Kepler) or newer can handle 4K.

And maybe try intermediate resolutions like 1440p? I presume 1440p monitors are pretty common nowadays.
Cryptocurrency has taught me to love math and at the same time be baffled by it.

Cryptocurrency lesson 0: Altcoins and Bitcoin are not the same thing.
 

Offline Howardlong

  • Super Contributor
  • ***
  • Posts: 4386
  • Country: gb
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #40 on: October 09, 2015, 06:20:58 pm »
I'm still not 100% convinced that these still image comparisons are valid, I mean, are we sure that we aren't grabbing an I frame from one video and a B frame from the other?

That's indeed a worthy observation, and one I hadn't considered.
 

Offline hans

  • Frequent Contributor
  • **
  • Posts: 998
  • Country: nl
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #41 on: October 09, 2015, 06:25:35 pm »
I'm still not 100% convinced that these still image comparisons are valid, I mean, are we sure that we aren't grabbing an I frame from one video and a B frame from the other? But based on other anecdotal evidence, we have to conclude that the 2160p30 viewed at 1080p30 is marginally better-looking than the 1080p30 native. I, for one, think 1080p50 looks far better than the indistuingishable (to me) 1080p30 and 2160p30. Even when it's just a talking head shot, it's just so fluid and life-like, I really like the 50fps.
Wouldn't grabbing frame 0 solve this issue? From my understanding you're basically talking about key-frames and intermediate frames.

I downloaded 1080p video, 4K upload at 1080p and 4K. Difference in file size is negligible at 1080p; so it's not a higher bitrate being pushed back to the clients. I did notice that the 4K and 1080p versions look very much alike in quality. The 1080p video has a lot more noise/artifacts in it.

Maybe the "raw" video is processed first to a format to be stored on YT servers, which needs obviously to be higher at 4K. Maybe from this intermediate file all other qualities are derived.
Because if you download your YouTube video after uploading, it is likely you won't get the same video file back, and often there is some quality loss.


I think Dave does a lot of "static" shots in his videos. Of course the mailbags and white-board talks are definitely not that and probably more pleasant to watch at 50fps, however a teardown often has static close-up images of the board (e.g. waffling on a topic for 2 minutes with 1 static board shot) that may be better viewed at a "higher quality" video.

But I don't think there is all that much in between them. I personally like to watch 50/60fps for any video with quick movement. I wouldn't have called the 1080p a bad encode if I hadn't view the 1080p @ 4K upload.
« Last Edit: October 09, 2015, 06:28:21 pm by hans »
 

Offline JuiceKing

  • Regular Contributor
  • *
  • Posts: 233
  • Country: us
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #42 on: October 09, 2015, 06:55:45 pm »
The 4K version certainly has more skin detail on my 27" monitor. It's pretty subtle and I didn't feel like I was missing out on anything with the regular HD. I suspect that lighting technique would have more impact on my overall impression of detail and sharpness.
 

Offline rs20

  • Super Contributor
  • ***
  • Posts: 2060
  • Country: au
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #43 on: October 09, 2015, 07:11:26 pm »
Wouldn't grabbing frame 0 solve this issue? From my understanding you're basically talking about key-frames and intermediate frames.

You're right about key frames vs intermediate frames; but I wouldn't recommend taking frame 0 because frame 0 is least likely to be representative of a typical frame (even if it were guaranteed to be a key frame).
 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 720
  • Country: se
  • Hobbyist
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #44 on: October 09, 2015, 11:59:00 pm »
I only have a 1920x1080 monitor. Both 2160 and 1080 looks the same to me, although in the posted still images in this thread there does indeed seem to be a small difference. 2160p stutters even though my computer and broadband should be fast enough (in theory), not sure why this is yet but it's not watchable as it is.

Wait, are you playing the 2160p video AT 2160p? Dave's OP specifically instructs us to play the 2160p video at 1080p:

Can you see any difference between this video at 1080p and this other one uploaded at 1080p?

If so, that's fascinating that the smaller file would be more demanding to decode AND higher quality to watch.
I only have a 1080p monitor so I have only been able to watch 2160p at 1080p. Smaller file-size, better quality and 4 times as many pixels is impressive. Ah, ok, thanks for pointing that out.
« Last Edit: October 10, 2015, 03:08:52 am by apis »
 

Offline rs20

  • Super Contributor
  • ***
  • Posts: 2060
  • Country: au
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #45 on: October 10, 2015, 12:02:02 am »
I only have a 1080p monitor so I have only been able to watch 2160p at 1080p. Smaller file-size, better quality and 4 times as many pixels is impressive.
You can still choose 2160p in YouTube even if you only have a 1080p screen, but I'm assuming you haven't done that.
 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 720
  • Country: se
  • Hobbyist
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #46 on: October 10, 2015, 12:25:59 am »
I only have a 1080p monitor so I have only been able to watch 2160p at 1080p. Smaller file-size, better quality and 4 times as many pixels is impressive.
You can still choose 2160p in YouTube even if you only have a 1080p screen, but I'm assuming you haven't done that.
I have tried it with the 2160p setting in youtube of course, that's when it stutters. Youtube scales it down to 1080p resolution though, since that's my monitors full screen resolution. I've watched the 2160p version at both 1080p and 2160p setting on youtube and compared both to the native 1080p EEVblog2 version (on my 1080p monitor (in full-screen mode of course)).
« Last Edit: October 10, 2015, 02:36:22 am by apis »
 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 720
  • Country: se
  • Hobbyist
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #47 on: October 10, 2015, 12:35:38 am »
If I pause it I can see small differences in quality but as I said before, I think 50/60fps vs 30fps makes a bigger difference: makes video look more sharp and fluid. And personally, since 2160p stutters, I can't use it anyway. But I can definitely manage with 1080p@30 fps as well so it's not a big deal either way. What's most important is that the original recording is in 50fps imho, can't really see much difference between the two videos.
« Last Edit: October 10, 2015, 03:03:09 am by apis »
 

Offline SteveLy

  • Regular Contributor
  • *
  • Posts: 220
  • Country: au
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #48 on: October 10, 2015, 01:45:04 am »
Quote
The 4K version ...
I think some posters are missing the point. You're meant to compare the two vids (EEVBlog and EEVBlog2 versions) at 1080p coming from youtube, not at 4K or at 4K from youtube scaled to a 1080p monitor.
« Last Edit: October 10, 2015, 01:47:26 am by SteveLy »
 

Offline apis

  • Frequent Contributor
  • **
  • Posts: 720
  • Country: se
  • Hobbyist
Re: eevBLAB #16 - 4K Render Youtube Experiment
« Reply #49 on: October 10, 2015, 02:58:29 am »
I'm starting to feel confused now :scared: but I've tried both settings and my conclusion is the same either way.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf