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

Video removed on YouTube ?

 

As a side note. Thanks for spending the time doing this Interceptor!

 

 

Sent from my iPhone using Tapatalk

Basically I have to keep it level 4.0 or 4.1 with fairly simple settings or it crashes

Either way youtube re-encoded it and whatever I was uploading it would accept

So the conclusion is so a file the way you can play typically level 4.0-4.1 with the slowest preset and let it go by itself

The standard output from imovie and final cut is really weak so you wonder why bother capturing at high quality if the encoding ruins it

Finally this is the 4K version of moses clips

This was the 1080p with x264

And this is the version using standard 1080p export

The theory is that 1080p downscaled should look better than the 4K downscaled by youtube or a native 1080p

I'm not sure what to look for in those downscaled 1080p clips as I'm viewing at home with a 4K TV so it's unfair to comment the result. The 4K however looks great. I'm looking at the sky and there is little artifact/noise compare to Adam's underwater clip. One thing can confirm is the LX100 lens beat the 12-42mm pz on the GH4, I'm glad I didn't go that route because the port 35 + gear would be $900 already (the price of the LX100).

When the weather get nice again, I will try to shoot a proper sky scene because I just got some ND filter, so 1/60 and F5.6 on the LX100 with proper exposed sky would be my next clip to try and with Handshake.

Thanks for doing all the testing!

For the 1080p check all clips in 1080p and compare to the 4K clip viewed at 1080p. I think some of the comparisons you are drawing are going to work on land but underwater it will be different as you can't shoot f/2 and you will miss the magnification of the 14-42mm when you try to do macro anyway that's another story

The weather is finally good enough for me to shoot with blue sky. This time I have a variable ND filter on so I can use the LX100 to shoot properly at 1/60, F5.6 (for maximize sharpness), and ISO200 (the lowest it would go). The variable ND filter I use is "Fotga", so it's not $300 kind of variable filter, but general consumer feedback on the variable ND filter is good so don't blame it on any artifact.

1.) Here is the first upload, the clip is right out of the camera, uploaded to Youtube. File size is 556Mb

http://youtu.be/I6KuRxYAZq4

2.) Here is the 2nd upload. The clip was brought in to Adobe Premiere CC, sequence setting is "Red Epic", during publish, the bit rate is set to target 80Mbps, and render is VBR 2 pass. File size is 453mb.

http://youtu.be/meur1gx5W5k

3.) This is the last upload. I took the 80Mbps clip output from Premiere CC and ran through Handbrake for Web Optimization. I will release the Handbrake setting info if people are interested. File size is 116Mb.

http://youtu.be/jHoFKwkDpY4

The reason of the test is to see which one Youtube reprocess and further compress the most. The wish is the 80Mbps version will be used for archive, and portfolio play back locally, so the bit rate is reasonably high at 80Mbps.

The Handbrake version is to hope that once the file is web optimized, Youtube would leave it alone and stream it as it is.

See which one you think is good and no artifact or blotch on the sky, which when come to underwater, the blue water is where we judge if the video is noisy or not.

Let me know if people are interested to download the original clip straight out of camera.

Last note, this is for 4K testing, so make sure to turn it on.

I am not sure about the options in premier but 2pass with H264 is really old school and so is target bitrate as you end up having issues with scene with lots of dynamic range or movement when you have sudden changes

Don't you have a constant quality option?

The first file is the best and this is just because it has only one conversion from original to youtube, the second has two and the third has three all of those introduce artifacts

You need to find the right settings in premiere as those you use are not great

By the way the camera seems to have quite a lot of rolling shutter issues

Are you referring to barrel distortion or rolling shutter? I don't see rolling shutter nor pan very fast. Regardless, rolling shutter shouldn't be an issue in the water since there is not straight line.

As far as setting for Premiere, I still haven't look into the optimal publish setting. So far I don't see the need of Handbrake as Youtube still process it despite the file is small. I don't know what could be done to trick Youtube to not process the file. Perhaps it's time to stop using Youtube and just use Vimeo?

Rolling shutter when panning. It's scatty not to the point of distorting but I can see it. The pan is too fast. From my tests see posts above both youtube and vimeo re-encode so the way forward is to produce the best possible file for your perusal and hope for the best. I used to produce two files for my rx100 and it looks like those don't get re-converted probably because sony doesn't use B frames at all. Panasonic instead does so the clips will always be-reconverted

The clear difference between youtube and vimeo is that for the same resolution vimeo manages higher bitrates so a file at the same resolution looks better on Vimeo

I briefly gone through some Youtube 4K diving videos just to check out what other camera has offered quality wise. To my surprise, there are quite a lot of Sony AX100 scuba footage, and the 4K quality seems better than the GH4 in most that I have seen. I'm not sure if it's the fact that Sony is using XAVC-S and Panasonic is using H.264. And that Youtube processed H.264 or leave videos that doesn't have B frame etc.

I thought this is interesting and see what we can investigate about other equipment result, and perhaps Panasonic's video get "discriminate" by Youtube and process the crap out of it.

Go check out other AX100 4K footage and see if you think the quality is sharper and blue water area has less banding:

Wow. That is interesting. This thread has been really interesting to read.

The thing I don't understand - if the GH4/Panasonic camera has these B frames. Would they not be discarded once you bring your GH4 footage into your NLE (as its editing in for example ProRes with FCPX) and then re-encoded by Premier or FCPX? Do these NLE also introduce B Frames?

Sent from my iPhone using Tapatalk

Sony ax100 is based on the same processor of the RX100 compact and records 4K at 60 Mbps which is 40% less than panasonic. Xavcs is also h264 so there is no difference there. Sony however tends to produce darker images with more contrast this actually reduces the issue. I had the same problem with the lx7 compared to the rx100

Wow. That is interesting. This thread has been really interesting to read.

 

The thing I don't understand - if the GH4/Panasonic camera has these B frames. Would they not be discarded once you bring your GH4 footage into your NLE (as its editing in for example ProRes with FCPX) and then re-encoded by Premier or FCPX? Do these NLE also introduce B Frames?

 

 

Sent from my iPhone using Tapatalk

The banding is produced when you record using soapy cinema like settings and had nothing to do with B frames. If you connect the camera to the tv or watch the clips as they are you will see it too before processing

I haven't had the chance to look at 4K scuba footage in Vimeo. If the quality in Vimeo is good among GH4/LX100 vs AX100. Then we can narrow it down that it's Youtube. Then from that point on, we just have to figure out what other 4K camera does that Panasonic doesn't do so that Youtube would give good quality. So far the thought is that Panasonic has B frame hence Youtube mess it up. If we can confirm that, then should we see if getting rid of B frame for Panasonic would improve the quality in Youtube?

I haven't had the chance to look at 4K scuba footage in Vimeo. If the quality in Vimeo is good among GH4/LX100 vs AX100. Then we can narrow it down that it's Youtube. Then from that point on, we just have to figure out what other 4K camera does that Panasonic doesn't do so that Youtube would give good quality. So far the thought is that Panasonic has B frame hence Youtube mess it up. If we can confirm that, then should we see if getting rid of B frame for Panasonic would improve the quality in Youtube?

Also the this clip is straight from the camera without grading and conversion that usually aggravate the issue

My LX100 test clip was out of the camera and yet the sky still has artifact. So clearly Youtube messed with it.

May be we should look for AX100 original somewhere and play with it? Like run it through FinalCut/Premiere then upload to Youtube and see if Youtube mess with it less then Panasonic's clips?

My LX100 test clip was out of the camera and yet the sky still has artifact. So clearly Youtube messed with it.

 

May be we should look for AX100 original somewhere and play with it? Like run it through FinalCut/Premiere then upload to Youtube and see if Youtube mess with it less then Panasonic's clips?

There is no banding in the files you gave me either on the files on youtube. In this case it has to do with the way you processed them I believe

Actually I don't see any banding in your upload from the camera either??

By the way the colors of that AX100 video are on the green side? It is very sharp though

Have read through this, it's a bit old but seem to have some quite good information, not sure what's IDR:

http://www.streamingmedia.com/Articles/Editorial/Featured-Articles/Encoding-for-YouTube-How-to-Get-the-Best-Results-83876.aspx

Also, I'm surprise that Youtube suggest 60 Mbps for 1080p video:

https://support.google.com/youtube/answer/1679498?hl=en&rd=1

I upload my 4K at 80Mbps. May be I should just publish my 4K at 100Mbps from now on and see what happen.

That article is good and still actual. The point at the start is what matters youtube will always re-encode. Concerning IDR it just means I frame the suggestion in the article is to use B frames or other frames as reference instead of increasing I frames which is fine. Also the youtube suggested GOP is already what panasonic does. What I don't understand Moses is that I see no banding in your 3 clips including handbrake only a bit of problem with the hazy sky but that's not banding??

The 60 Mbps suggestion 4:2:2 is for tv broadcast style it doesn't apply to us common mortals

There is not banding on the 3 video I uploaded recently, but there are some artifacts. I will see if I can do a screen capture when I get home.

The article also suggest increase the number of B-frame, "like B-frame interval of 2". But I still don't know how it affect the result because the video output from Youtube doesn't have B-frame.

Artefacts always occur when you re-encode that's why you should try and edit in native format if your computer allows as every conversion is lossy. Nothing new here??