Getting the best results from YouTube / Vimeo (1080, 4K etc...)

Thought I would start the ball rolling on this one as there doesn't seem to appear to be a one size fits all answer to this...

"what are the best settings / encoder to use, to get the best possible results when I upload a movie to YouTube or Vimeo?"

I don't know the answer so would be great to get any opinions or solutions people have used before. :-)

Someone on RedUser has already spent some time looking at it

http://www.reduser.net/forum/showthread.php?70894-YouTube-Video-Codecs-Resolutions-amp-Data-Rates&p=924139&viewfull=1#post924139

But maybe with underwater footage, the outcome would be different.

Sent from my iPhone using Tapatalk

That post is out dated referencing youtube at 2-5 mbps

Uploading video in pro-res format is not only not needed but also takes forever considering the size ratio with h264 is 10:1 and the is no quality improvement perceived

The reference is here

Youtube

https://support.google.com/youtube/answer/1722171?hl=en-GB

Vimeo

https://vimeo.com/help/compression

Note that vimeo supports much higher bitrate for the same format however you need to upgrade quickly as the limit of 500 Mb/week is easily used up

If you have a PRO account there is no bitrate issue otherwise youtube will re-encode

The issue is that programs like final cut X use an inefficient h264 codec that does not make the most of the bitrate however with legacy versions of final cut and legacy versions of iMovie you can use x264 with full encode control

I do not know about windows program so can't comment but the issue is taking control of the output codec settings

Thanks for the link. Very interesting!!

Loads of detail!!

Nothing about VP9 which is weird thou. You would think they would say something about it.

What's x264 - new one on me.

There surely must be people out there who do this sort of thing professionally (YouTube uploading etc) who know what's the secret source!

Too be honest I've only using the built-in 'Send To... YouTube' feature of FCPX. As I presumed it would provide the best settings.

Sent from my iPhone using Tapatalk

Learning a LOT today!

So x264 is an open source encoder of h264 - with massively better results.

Good write up here - http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/How-to-Produce-High-Quality-H.264-Video-Files-94216.aspx

As to using an old version of iMovie. I've just read that you could export a ProRes file and then use the free Handbrake app to encode to h264 - BUT via x264 as it uses that as its h264 encoder.

Sent from my iPhone using Tapatalk

x264 is the best mac plug in to encode h264 in quick time. There are light years of difference between x264 high profile encoding with very slow presets and a youtube export out of the box

The plug in is on git https://github.com/MyCometG3/x264Encoder/

Example export of iMovie for youtube 645 MB

Video

ID : 1

Format : AVC

Format/Info : Advanced Video Codec

Format profile : Main@L4.1

Format settings, CABAC : No

Format settings, ReFrames : 2 frames

Format settings, GOP : M=2, N=50

Codec ID : avc1

Codec ID/Info : Advanced Video Coding

Duration : 4mn 24s

Bit rate : 20.3 Mbps

Width : 1 920 pixels

Height : 1 080 pixels

Display aspect ratio : 16:9

Frame rate mode : Constant

Frame rate : 25.000 fps

Color space : YUV

Chroma subsampling : 4:2:0

Bit depth : 8 bits

Scan type : Progressive

Bits/(Pixel*Frame) : 0.392

Stream size : 640 MiB (99%)

Language : English

Encoded date : UTC 2015-01-24 18:57:46

Tagged date : UTC 2015-01-24 19:08:52

Color primaries : BT.709

Transfer characteristics : BT.709

Matrix coefficients : BT.709

The export is not actually optimised for youtube and has a too high bitrate CABAC off means the file is huge the profile is main quality is low and youtube will re-encode it making it worse

Exporting the same using the option export to quicktime and setting X264 options correctly will result in a compliant file that youtube will not convert that will actually look better

The secret is to avoid youtube doing its own thing

Both iMovie Final Cut can edit native X264 once they are in the correct container and again if the bitrate is too low (24 Mbps 1080p or 100 mbps 4K) grading is not an option

If you start instead with 100 mbps 1080p file you can produce a very nice file for yourself but unless you have a youtube pro account the end result will be average

Most of the 4K clips I have seen around suffer from too much workflow applied and conversion so in the end they look soft

If someone has a the source of the nauticam LX100 and a GH4 high quality file on dropbox I can play them with my Minix on my 4K 55" screen to see how far are they from 1080p

When I play my original 1080p and compare to high quality 4K on the TV only the backscatter example beats it even ejunae files look discoloured

You can download both GH4 and LX100 footage here (they're straight out of the camera) and using Scenery color profile.

1.) LX100 shot at Manual mode using F5.6 (the sharpest for the lens), ISO200 (the lowest the camera can go, ISO100 is only for photo), Shutter speed 1/1300.

2.) LX100 shot at Manual mode using F5.6, ISO200, Shutter speed 1/1600 (to under expose a bit and get a deeper blue.

3.) LX100 shot at Shutter priority mode, ISO200, Shutter speed 1/60.

4.) GH4 + 14-140mm shot at Manual mode using F5.6, ISO200, Shutter speed 1/1600 to under expose a bit and get deeper blue. This one is using the Red dedicated movie button to start recording.

5.) Same as (4) but using the shutter button to start recording, don't think there is a difference whether you use the shutter button or the red button to start recording.

6.) GH4+ 14-140mm shot at Shutter priority mode, ISO200, Shutter speed 1/60. using MP4 mode instead of MOV to compare with LX100

1.) has some hunting issue so I turned off Continuous focus to shoot the rest. Today is the first day when I have sunny day during winter since I have the LX100. I'm struggle with over expose at 1/60. I don't know what is the other acceptable shutter speed to use for 30fps. I tried 1/1300 and 1/1600, I get the exposure I wanted, but the side effect is there is flicker in the middle of the screen. I could really use an ND filter for the LX100!

Feel free to download and play with the clip. I think I'm the only non Mac user here. Here are my workflow:

1.) load clips into Adobe Premiere CC to compose and arrange clips

2.) Export Final Cut Pro XML from Premiere

3.) Using Davinci Resolve to import the Final Cut Pro XML and do color grading.

4.) Export graded video from Davinci to 10-bit Uncompress RGB with Final Cut Pro XML

5.) Use Premiere CC to open the Final Cut Pro XML from Davinci, add music and output to H.264 80Mbps.

I don't care about all those 100Mbps original shouldn't be graded comment.To me, there is a need of grading from time to time since Custom White Balance could be a bit off depend on the angle the white slate is held etc. Besides, Many Indie file makers who use GH4 use the workflow I describe above to use Davinci to grade and rountrip to Premiere. If it's good enough for them, I don't see any issue of using it, especially the graded video looks good on my local computer/tablet as portfolio.

BTW, read that the Roast Duck 4K video in the other discussion was shot with RED.

Some info on 4k to 1080p downscaling..

http://www.shutterangle.com/2014/shooting-4k-video-for-2k-delivery-bitdepth-advantage/

Explained it really well!

Moses for land use you need an ND8 filter to put on the thread of the camera. It is an essential accessory to have for any land video shooting

At 30p you want to be shooting at 1/60 ideally or 1/125 without aperture too small

In water there won't be such problem

When I said grading is not worth it I meant in water, on land you can do a bit of it and this is the example you refer to

The other example from Richard assumes you can use 4:4:4 chroma sampling but none of youtube or vimeo use anything different than 4:2:0, the LX100 I am not even sure it can take an external recorder

So all this stuff you see 'pro' people do is because the shoot it for their own projection or stream directly not using youtube or other cheapo services that at the very end finish with 4:2:0

Now downscaling to 1080p will give you more grading ability because of the 4:4:4 is very similar to editing adobe RGB RAW files you have more latitude to play with

However you need enough bitrate to start with and scaling is not done so amazingly as you think in software

Ok, I think I'm getting it. Read that top side videographers do use ND filter, specifically "Variable ND filter". Most Variable ND filters soften the images. But for now, I think there is a bit of advantage for underwater. For the LX100 short port, I will be using a red filter, which will cut down the light by 2-stop. That should help a bit. The problem is, not sure what to do when shooting into the sky for sunball or dapple light scene.

You can get a variable ND filter if the thread on the LX100 lens is big enough. If the thread is small an ND8 which is actually 3 stops is what you need. Now in water you don't really want to use an ND filter and you can't shoot sunballs with a filter as you get red tint. Generally you will use small apertures as opposed to fast shutter speed otherwise the water ripples will look scatty. I think the LX100 smallest aperture is f/16 this may still require shutter of 1/250 at ISO 200 but you really don't want to go in the thousands. I downloaded two of your test files am going to look into them later today

Those are the two files

LX100

Video

ID : 1

Format : AVC

Format/Info : Advanced Video Codec

Format profile : High@L5.1

Format settings, CABAC : Yes

Format settings, ReFrames : 2 frames

Format settings, GOP : M=3, N=8

Codec ID : avc1

Codec ID/Info : Advanced Video Coding

Duration : 38s 38ms

Bit rate mode : Variable

Bit rate : 95.8 Mbps

Maximum bit rate : 106 Mbps

Width : 3 840 pixels

Height : 2 160 pixels

Display aspect ratio : 16:9

Frame rate mode : Constant

Frame rate : 29.970 fps

Color space : YUV

Chroma subsampling : 4:2:0

Bit depth : 8 bits

Scan type : Progressive

Bits/(Pixel*Frame) : 0.385

Stream size : 434 MiB (100%)

Encoded date : UTC 2015-01-24 13:55:27

Tagged date : UTC 2015-01-24 13:55:27

Color primaries : BT.709

Transfer characteristics : BT.709

Matrix coefficients : BT.709

ID : 1

Format : AVC

Format/Info : Advanced Video Codec

Format profile : High@L5.1

Format settings, CABAC : Yes

Format settings, ReFrames : 2 frames

Format settings, GOP : M=3, N=8

Codec ID : avc1

Codec ID/Info : Advanced Video Coding

Duration : 31s 532ms

Bit rate mode : Variable

Bit rate : 79.5 Mbps

Maximum bit rate : 106 Mbps

Width : 3 840 pixels

Height : 2 160 pixels

Display aspect ratio : 16:9

Frame rate mode : Constant

Frame rate : 29.970 fps

Color space : YUV

Chroma subsampling : 4:2:0

Bit depth : 8 bits

Scan type : Progressive

Bits/(Pixel*Frame) : 0.320

Stream size : 299 MiB (100%)

Encoded date : UTC 2015-01-24 15:01:39

Tagged date : UTC 2015-01-24 15:01:39

Color primaries : BT.709

Transfer characteristics : BT.709

Matrix coefficients : BT.709

Randy (MrScuba) is getting the dome port. For me, $280 for the dome port just to restore air FOV is not useful for me. Feel free to upload the clip to Youtube/Vimeo and see what Youtube does to the video. I will be busy for the next few weeks to try out the LX100 with my setup.

If you were able to restore the full air field of view the dome would be extremely useful because on land you really don't miss much and in water if you go for a normal reef dive it is all you need. Also I think you can still use a wet lens using an M67 arm though this looks a bit on the edge

I will see what happens to those files but the essence is that the codec and processing is the same. The difference between the camera is the lenses and the sensor not the processing so in theory you should have very similar performance at comparable focal length 24-75mm.

I will see if I can squeeze those files and see how to get something compliant

Great thread

Adam

Am having some challenges as prores gets to 220 mbits all I frames but it's only 4:2:0 so am not able to implement the theory of the post from trickster. I will anyway try and give it a go to see what survives after scaling. Looks like 4K is a bit beyond the current codec offering or it needs much higher bitrates for extensive manipulation

What's your work flow? May be some other people here might try the same. I'm on PC and the Premiere <-> Davinci round trip is the only thing I do now. And I haven't try it on my i7 Laptop to see if it's doable. On my desktop, I have Nvidia card so the CUDA features in Premiere makes speed doable.

I'm using FCPX on Yosemite on either a i7 rMBP or a i5 Air - both handle 4k footage (obviously the rMBP is much faster!)

I import my clips (create proxies, but not optimised proress files) mess with them (with FCPX running in a 444 32bit workspace) and then use the builtin 'Send to...YouTube' function.

Ok I found how to do it. You install x264 and apple compressor and then when you go to advanced settings you can fully play with the profiles. However the issue is that if I do color grading in final cut before scaling the colour space is 4:2:0 so there is not a lot to play with. Solution you need an external recorder like shogun to capture 4K 4:2:2 or you need to scale first to then be able to operate in 4444. I am now trying the second option as it is the only available to those not having a shogun and currently there is no housing for it. So my workflow is mov to prores 422 hq or 4444xq then x264. Ideally you want 4k 422 to x264

However the issue is that if I do color grading in final cut before scaling the colour space is 4:2:0 so there is not a lot to play with.

When you place a clip on the timeline within FCPX - all effects to that clip are in 444 32bit - and at the native resolution of the clip

(no need to optimise the clip to ProRes - as its doing that in the background)

However pre-optimizing your clips might make general use of FCPX quicker as its not doing the background rendering as much.

Thats what I understand anyway...

Am not convinced something becomes something else with nothing done to it. Your articles suggests you need to scale to get 444 if this was possible without scaling why would you bother? Unless the setting you mention works in both a 444 color space realtime and downscales which looks a bit challenging as there would not be a fixed basis for color adjustment on a moving target. Why would devices like shogun be sold if this could have been worked out in software? By recording 4:2:0 you are cutting out information you need to scale to get the colour information sufficient to sustain extensive edit at least this is what I understood from your previous link